Download presets_data.js from pavi-07/satquery: direct link, hf CLI and curl.
- Browser
- Download file 82.9 kB
-
https://huggingface.co/spaces/pavi-07/satquery/resolve/main/presets_data.js
- Command line
-
hf download hf://spaces/pavi-07/satquery/presets_data.js
-
curl -L -o presets_data.js https://huggingface.co/spaces/pavi-07/satquery/resolve/main/presets_data.js
82.9 kB
| window.SATQUERY_PRESETS = { | |
| "fusion": { | |
| "answer": "### \ud83c\udf0a Geospatial Intelligence Briefing: Flood Inundation Assessment\n\n**Target AOI:** Brahmaputra River Basin, Assam, India\n**Sensor Telemetry:** Sentinel-1 C-Band SAR (VV Polarization) + Sentinel-2 MSI Multi-Spectral\n**Cloud Penetration Status:** \u2705 **100% Surface Penetration** (Zero attenuation through monsoonal cloud cover)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nStandard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.\n\n#### \ud83c\udfa8 Visual Interpretation & Color Guide\n\u2022 \ud83d\udd34 **Bright Red / Orange:** High neural activation \u2014 **Active standing floodwater** (specular radar reflection, zero backscatter).\n\u2022 \ud83d\udfe1 **Yellow / Light Green:** Moderate activation \u2014 **Saturated ground, wetlands, and waterlogged cropland**.\n\u2022 \ud83d\udd35 **Blue / Dark Cyan:** Low activation \u2014 **Dry, elevated ground & unaffected terrain** (safe zones).\n\ud83d\udca1 *Interactive Tip: Use the buttons above the viewer to toggle between **Sentinel-2 Optical** (shows monsoonal clouds), **Sentinel-1 SAR** (shows radar surface penetration), and **Grad-CAM Heatmap** (shows AI class activation).* \n\n#### \ud83d\udef0\ufe0f Multi-Modal Findings & Analysis\n\u2022 **Inundated / Surface Water:** **91% Calibrated Confidence** \u2014 Confirmed active water inundation across riverine corridors.\n\u2022 **Radar vs. Optical Penetration:** Optical bands encounter dense cloud cover scattering; C-band synthetic aperture radar penetrates 100% of atmospheric moisture, detecting characteristic low-dielectric specular backscatter of standing water.\n\u2022 **Shoreline & Agricultural Interaction:** Strong radar contrast demarcates submerged cropland and compromised embankment sections.\n\n#### \u26a0\ufe0f Risk Classification & Recommended Advisory\n\u2022 **Status:** **STAGE-2 INUNDATION WARNING (RED ADVISORY)**\n\u2022 **Action:** Immediate hydrological monitoring along the alluvial floodplain. Alert local district disaster management authorities. Prioritize emergency staging in elevated blue zones.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udf0a Geospatial Intelligence: Flood Inundation Assessment</div>\n <div class=\"report-subtitle\">Target AOI: <b>Brahmaputra River Basin, Assam, India</b> | Cloud Penetration: <b>100% Surface Penetration</b> | Sensors: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>Standard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: Active flood threat in progress. Emergency response teams should focus on red/yellow sectors.</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Read the Imagery & Heatmap</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#ef4444;\"></span> <div><b>Bright Red / Orange:</b> Active standing floodwater (highest hazard zone).</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#eab308;\"></span> <div><b>Yellow / Green:</b> Saturated, waterlogged soil, agricultural runoff & marshes.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Blue / Dark Cyan:</b> Dry elevated ground & unaffected terrain (safe zones).</div></div>\n <div style=\"font-size:11.5px; color:#64748b; margin-top:6px; font-style:italic;\">Tip: Use the buttons above the image to switch between Optical (cloudy), SAR Radar (penetrated), and the Heatmap.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udef0\ufe0f Multi-Modal Findings & Analysis</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Inundated Surface Water:</b> <b>91% Calibrated Confidence</b> \u2014 Confirmed active water inundation across riverine corridors.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Radar Penetration:</b> Optical bands blocked by monsoonal clouds; C-band synthetic aperture radar achieves 100% cloud penetration, identifying specular water reflection.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Agricultural Impact:</b> Sharp radar contrast isolates submerged paddy fields and compromised embankment perimeters.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Risk Classification & Advisory</b>\n <p><b>STAGE-2 INUNDATION WARNING (RED ADVISORY):</b> Continue regular hydrological monitoring across low-lying floodplain sectors. Emergency services should maintain readiness in designated elevated zones.</p>\n </div>\n </div>\n ", | |
| "confidence": 0.9091, | |
| "confidence_label": "Confidence: 91% (Calibrated Sigmoid)", | |
| "confidence_note": "Calibrated multi-label probability from Optical-SAR dual-branch CNN", | |
| "confidence_percent": 91, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.96, | |
| "date_or_time_reference": null, | |
| "domain": "disaster", | |
| "location": "Assam", | |
| "needs_clarification": false, | |
| "raw_query": "Combine optical and SAR imagery to identify flooded areas in Assam during cloud cover.", | |
| "sensor": "both", | |
| "source": "llm", | |
| "target_object": null, | |
| "task_sequence": [ | |
| "fusion_analysis" | |
| ] | |
| }, | |
| "is_compound": false, | |
| "is_confidence_reliable": true, | |
| "load_events": [], | |
| "location": "Brahmaputra River Basin, Assam, India", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Brahmaputra River Basin, Assam, India", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "optical": "assets/assam_optical_rgb.png", | |
| "sar": "assets/assam_sar_radar.png" | |
| }, | |
| "query": "Combine optical and SAR imagery to identify flooded areas in Assam during cloud cover.", | |
| "results": [ | |
| { | |
| "answer": "### \ud83c\udf0a Geospatial Intelligence Briefing: Flood Inundation Assessment\n\n**Target AOI:** Brahmaputra River Basin, Assam, India\n**Sensor Telemetry:** Sentinel-1 C-Band SAR (VV Polarization) + Sentinel-2 MSI Multi-Spectral\n**Cloud Penetration Status:** \u2705 **100% Surface Penetration** (Zero attenuation through monsoonal cloud cover)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nStandard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.\n\n#### \ud83c\udfa8 Visual Interpretation & Color Guide\n\u2022 \ud83d\udd34 **Bright Red / Orange:** High neural activation \u2014 **Active standing floodwater** (specular radar reflection, zero backscatter).\n\u2022 \ud83d\udfe1 **Yellow / Light Green:** Moderate activation \u2014 **Saturated ground, wetlands, and waterlogged cropland**.\n\u2022 \ud83d\udd35 **Blue / Dark Cyan:** Low activation \u2014 **Dry, elevated ground & unaffected terrain** (safe zones).\n\ud83d\udca1 *Interactive Tip: Use the buttons above the viewer to toggle between **Sentinel-2 Optical** (shows monsoonal clouds), **Sentinel-1 SAR** (shows radar surface penetration), and **Grad-CAM Heatmap** (shows AI class activation).* \n\n#### \ud83d\udef0\ufe0f Multi-Modal Findings & Analysis\n\u2022 **Inundated / Surface Water:** **91% Calibrated Confidence** \u2014 Confirmed active water inundation across riverine corridors.\n\u2022 **Radar vs. Optical Penetration:** Optical bands encounter dense cloud cover scattering; C-band synthetic aperture radar penetrates 100% of atmospheric moisture, detecting characteristic low-dielectric specular backscatter of standing water.\n\u2022 **Shoreline & Agricultural Interaction:** Strong radar contrast demarcates submerged cropland and compromised embankment sections.\n\n#### \u26a0\ufe0f Risk Classification & Recommended Advisory\n\u2022 **Status:** **STAGE-2 INUNDATION WARNING (RED ADVISORY)**\n\u2022 **Action:** Immediate hydrological monitoring along the alluvial floodplain. Alert local district disaster management authorities. Prioritize emergency staging in elevated blue zones.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udf0a Geospatial Intelligence: Flood Inundation Assessment</div>\n <div class=\"report-subtitle\">Target AOI: <b>Brahmaputra River Basin, Assam, India</b> | Cloud Penetration: <b>100% Surface Penetration</b> | Sensors: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>Standard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: Active flood threat in progress. Emergency response teams should focus on red/yellow sectors.</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Read the Imagery & Heatmap</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#ef4444;\"></span> <div><b>Bright Red / Orange:</b> Active standing floodwater (highest hazard zone).</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#eab308;\"></span> <div><b>Yellow / Green:</b> Saturated, waterlogged soil, agricultural runoff & marshes.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Blue / Dark Cyan:</b> Dry elevated ground & unaffected terrain (safe zones).</div></div>\n <div style=\"font-size:11.5px; color:#64748b; margin-top:6px; font-style:italic;\">Tip: Use the buttons above the image to switch between Optical (cloudy), SAR Radar (penetrated), and the Heatmap.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udef0\ufe0f Multi-Modal Findings & Analysis</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Inundated Surface Water:</b> <b>91% Calibrated Confidence</b> \u2014 Confirmed active water inundation across riverine corridors.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Radar Penetration:</b> Optical bands blocked by monsoonal clouds; C-band synthetic aperture radar achieves 100% cloud penetration, identifying specular water reflection.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Agricultural Impact:</b> Sharp radar contrast isolates submerged paddy fields and compromised embankment perimeters.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Risk Classification & Advisory</b>\n <p><b>STAGE-2 INUNDATION WARNING (RED ADVISORY):</b> Continue regular hydrological monitoring across low-lying floodplain sectors. Emergency services should maintain readiness in designated elevated zones.</p>\n </div>\n </div>\n ", | |
| "confidence": 0.9091, | |
| "confidence_label": "Confidence: 91% (Calibrated Sigmoid)", | |
| "confidence_note": "Calibrated multi-label probability from Optical-SAR dual-branch CNN", | |
| "confidence_percent": 91, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Optical-SAR Fusion", | |
| "is_confidence_reliable": true, | |
| "specialist": "fusion", | |
| "step": 1, | |
| "task_type": "fusion_analysis", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "fusion", | |
| "specialist_sequence": [ | |
| "fusion" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "fusion_analysis" | |
| ], | |
| "task_type": "fusion_analysis", | |
| "top_classes": [] | |
| }, | |
| "change_vqa": { | |
| "answer": "### \ud83d\udd04 Bi-Temporal Change Detection: Observational Discrepancy Flagged\n\n**Target Corridor:** Metropolitan Growth Corridor\n**Observation Windows:** T\u2081 (Pre-Disaster / Baseline) \u2194 T\u2082 (Post-Disaster / Present)\n**Radiometric Ground Telemetry:** \u26a0\ufe0f **74.1% Measured Surface Variance** Detected\n\n#### \u26a0\ufe0f Ambiguity & Observational Discrepancy Alert\n\u2022 **Model Inference:** The vision-language model predicted *'No verified completed building envelopes'*.\n\u2022 **Radiometric Reality:** Physical pixel telemetry measures **74.1% surface modification** between observation dates.\n\u2022 **Model Limitation Note:** Greedy token decoding certainty (~100%) is uncalibrated. Benchmark accuracy is ~72.5%.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nWhile the vision model did not definitively confirm completed rectangular building rooftops, **noticeable ground disturbances, site grading, or surface alterations are clearly visible** in the satellite scene (74.1% changed area). **Do NOT treat this scene as unchanged.** Active land preparation, earth clearing, or environmental conversion is taking place.\n\n#### \ud83c\udfa8 How to Interpret the Comparison Slider\n\u2022 \u23ea **Left Image (Before):** Historical baseline state.\n\u2022 \u23e9 **Right Image (After):** Recent satellite capture showing altered surface.\n\u2022 \u2194 **Mandatory Visual Inspection:** Drag the slider handle across the scene to manually inspect the altered boundaries and ground clearings.\n\n#### \ud83d\udcca Comparative Analytical Findings\n\u2022 **Structural Rooftop Status:** Negative / Unconfirmed by VLM.\n\u2022 **Physical Surface Modification:** **74.1% Altered Land Surface** (Confirmed by radiometric delta).\n\u2022 **Site Condition:** **Active Modification in Progress** (Land cleared / earthworks observed).\n\n#### \u26a0\ufe0f Operational Advisory\n\u2022 **Status:** **AMBER ADVISORY: UNVERIFIED SURFACE MODIFICATION**\n\u2022 **Action:** Flagged for mandatory human inspection. Do not approve as structurally static without site verification.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\" style=\"color:#b91c1c;\">\u26a0\ufe0f Bi-Temporal Change Detection: Observational Discrepancy Flagged</div>\n <div class=\"report-subtitle\">Target Corridor: <b>Metropolitan Growth Corridor</b> | Radiometric Telemetry: <b>74.1% Surface Variance</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"discrepancy-box\">\n <b>\u26a0\ufe0f Observational Discrepancy & Ambiguity Alert</b>\n <p>The vision model predicted <i>\"no verified building envelopes\"</i>, but radiometric pixel analysis detects <b>74.1% physical surface variance</b> between T\u2081 and T\u2082. Visible land grading, surface clearance, or environmental modification has occurred.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#9f1239; font-weight:700;\">Flagged: Do NOT treat this area as static. Active surface disturbance detected.</div>\n </div>\n\n <div class=\"simple-words-box\" style=\"background:#fffbeb; border-color:#f59e0b;\">\n <b style=\"color:#92400e;\">\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>While the AI could not definitively classify completed rectangular building rooftops, <b>clear physical changes and earthworks have altered the ground</b> across 74.1% of the monitored scene. Preparatory construction or ground development is in progress.</p>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Use the Comparison Slider Above</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#059669;\"></span> <div><b>Left Image (Before):</b> Historical baseline reference.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#e11d48;\"></span> <div><b>Right Image (After):</b> Recent satellite capture displaying altered ground.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#6366f1;\"></span> <div><b>Drag Slider (\u2194):</b> Move handle to visually verify surface grading and cleared parcel boundaries.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udcca Comparative Analytical Findings</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Rooftop Classification:</b> Unconfirmed by VLM (Model returned negative token).</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Physical Surface Delta:</b> <b>74.1% Altered Land Surface</b> detected via radiometric subtraction.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Site Condition:</b> <b>Active Ground Modification</b> \u2014 Land clearing or earthwork in progress.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Operational Advisory: UNVERIFIED SURFACE MODIFICATION</b>\n <div>Flagged for mandatory manual review. Do not issue a static clearance certificate without field inspection of altered boundaries.</div>\n </div>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Discrepancy Flagged (Pixel Variance Detected)", | |
| "confidence_note": "Semantic model predicted no change, but radiometric delta detected 74.1% surface modification", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.95, | |
| "date_or_time_reference": "before and after", | |
| "domain": "general", | |
| "location": null, | |
| "needs_clarification": false, | |
| "raw_query": "Did the buildings change between the before and after images?", | |
| "sensor": "unspecified", | |
| "source": "llm", | |
| "target_object": null, | |
| "task_sequence": [ | |
| "change_vqa" | |
| ] | |
| }, | |
| "is_compound": false, | |
| "is_confidence_reliable": false, | |
| "load_events": [], | |
| "location": "Metropolitan Growth Corridor", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Metropolitan Growth Corridor", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "after": "assets/second_after_00857.png", | |
| "before": "assets/second_before_00857.png" | |
| }, | |
| "query": "Did the buildings change between the before and after images?", | |
| "results": [ | |
| { | |
| "answer": "### \ud83d\udd04 Bi-Temporal Change Detection: Observational Discrepancy Flagged\n\n**Target Corridor:** Metropolitan Growth Corridor\n**Observation Windows:** T\u2081 (Pre-Disaster / Baseline) \u2194 T\u2082 (Post-Disaster / Present)\n**Radiometric Ground Telemetry:** \u26a0\ufe0f **74.1% Measured Surface Variance** Detected\n\n#### \u26a0\ufe0f Ambiguity & Observational Discrepancy Alert\n\u2022 **Model Inference:** The vision-language model predicted *'No verified completed building envelopes'*.\n\u2022 **Radiometric Reality:** Physical pixel telemetry measures **74.1% surface modification** between observation dates.\n\u2022 **Model Limitation Note:** Greedy token decoding certainty (~100%) is uncalibrated. Benchmark accuracy is ~72.5%.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nWhile the vision model did not definitively confirm completed rectangular building rooftops, **noticeable ground disturbances, site grading, or surface alterations are clearly visible** in the satellite scene (74.1% changed area). **Do NOT treat this scene as unchanged.** Active land preparation, earth clearing, or environmental conversion is taking place.\n\n#### \ud83c\udfa8 How to Interpret the Comparison Slider\n\u2022 \u23ea **Left Image (Before):** Historical baseline state.\n\u2022 \u23e9 **Right Image (After):** Recent satellite capture showing altered surface.\n\u2022 \u2194 **Mandatory Visual Inspection:** Drag the slider handle across the scene to manually inspect the altered boundaries and ground clearings.\n\n#### \ud83d\udcca Comparative Analytical Findings\n\u2022 **Structural Rooftop Status:** Negative / Unconfirmed by VLM.\n\u2022 **Physical Surface Modification:** **74.1% Altered Land Surface** (Confirmed by radiometric delta).\n\u2022 **Site Condition:** **Active Modification in Progress** (Land cleared / earthworks observed).\n\n#### \u26a0\ufe0f Operational Advisory\n\u2022 **Status:** **AMBER ADVISORY: UNVERIFIED SURFACE MODIFICATION**\n\u2022 **Action:** Flagged for mandatory human inspection. Do not approve as structurally static without site verification.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\" style=\"color:#b91c1c;\">\u26a0\ufe0f Bi-Temporal Change Detection: Observational Discrepancy Flagged</div>\n <div class=\"report-subtitle\">Target Corridor: <b>Metropolitan Growth Corridor</b> | Radiometric Telemetry: <b>74.1% Surface Variance</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"discrepancy-box\">\n <b>\u26a0\ufe0f Observational Discrepancy & Ambiguity Alert</b>\n <p>The vision model predicted <i>\"no verified building envelopes\"</i>, but radiometric pixel analysis detects <b>74.1% physical surface variance</b> between T\u2081 and T\u2082. Visible land grading, surface clearance, or environmental modification has occurred.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#9f1239; font-weight:700;\">Flagged: Do NOT treat this area as static. Active surface disturbance detected.</div>\n </div>\n\n <div class=\"simple-words-box\" style=\"background:#fffbeb; border-color:#f59e0b;\">\n <b style=\"color:#92400e;\">\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>While the AI could not definitively classify completed rectangular building rooftops, <b>clear physical changes and earthworks have altered the ground</b> across 74.1% of the monitored scene. Preparatory construction or ground development is in progress.</p>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Use the Comparison Slider Above</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#059669;\"></span> <div><b>Left Image (Before):</b> Historical baseline reference.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#e11d48;\"></span> <div><b>Right Image (After):</b> Recent satellite capture displaying altered ground.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#6366f1;\"></span> <div><b>Drag Slider (\u2194):</b> Move handle to visually verify surface grading and cleared parcel boundaries.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udcca Comparative Analytical Findings</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Rooftop Classification:</b> Unconfirmed by VLM (Model returned negative token).</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Physical Surface Delta:</b> <b>74.1% Altered Land Surface</b> detected via radiometric subtraction.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Site Condition:</b> <b>Active Ground Modification</b> \u2014 Land clearing or earthwork in progress.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Operational Advisory: UNVERIFIED SURFACE MODIFICATION</b>\n <div>Flagged for mandatory manual review. Do not issue a static clearance certificate without field inspection of altered boundaries.</div>\n </div>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Discrepancy Flagged (Pixel Variance Detected)", | |
| "confidence_note": "Semantic model predicted no change, but radiometric delta detected 74.1% surface modification", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Change Detection", | |
| "is_confidence_reliable": false, | |
| "specialist": "qwen_cdvqa", | |
| "step": 1, | |
| "task_type": "change_vqa", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "qwen_cdvqa", | |
| "specialist_sequence": [ | |
| "qwen_cdvqa" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "change_vqa" | |
| ], | |
| "task_type": "change_vqa", | |
| "top_classes": [] | |
| }, | |
| "vqa": { | |
| "answer": "### \ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report\n\n**Target AOI:** Chennai, Tamil Nadu, India\n**Platform / Sensor:** High-Resolution Optical Satellite Telemetry (RGB)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe land cover around Chennai appears to be predominantly agricultural, with fields and crops visible The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.\n\n#### \ud83c\udfa8 Visual Guide \u2014 How to Read This Scene\n\u2022 **Dark Linear Corridors:** Paved highways, local access roads, and rail corridors.\n\u2022 **Geometric Blocks:** Commercial and residential building rooftops and structures.\n\u2022 **Green / Textured Patches:** Agricultural plots, tree foliage, and open permeable ground.\n\n#### \ud83d\udccb Analytical Assessment\nThe land cover around Chennai appears to be predominantly agricultural, with fields and crops visible. The area is likely used for farming or other agricultural purposes.\n\n#### \ud83d\udd0d Sensor & Model Telemetry\n\u2022 **Sensor Telemetry:** Multi-spectral visual resolution capturing physical ground contours.\n\u2022 **Inference Architecture:** Qwen2.5-VL vision-language model with multi-scale vision transformer.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report</div>\n <div class=\"report-subtitle\">Target AOI: <b>Chennai, Tamil Nadu, India</b> | Model: <b>Qwen2.5-VL</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The land cover around Chennai appears to be predominantly agricultural, with fields and crops visible The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Direct Answer to: \"What does the land cover look like around Chennai?\"</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 Visual Guide \u2014 How to Read This Scene</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#334155;\"></span> <div><b>Dark Linear Corridors:</b> Paved highways, local access roads, and rail corridors.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Geometric Rectangles:</b> Residential and commercial building rooftops.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#16a34a;\"></span> <div><b>Textured Patches:</b> Agricultural fields, green foliage, and open permeable soil.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Comprehensive Analytical Assessment</div>\n <div style=\"font-size:13px; line-height:1.6; color:#2c4b50; margin-bottom:12px; background:#f8fafb; padding:12px 14px; border-radius:7px; border:1px solid #e2e8ea;\">\n The land cover around Chennai appears to be predominantly agricultural, with fields and crops visible. The area is likely used for farming or other agricultural purposes.\n </div>\n\n <div class=\"report-section-title\">\ud83d\udd0d Telemetry & Sensor Parameters</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Ground Sampling:</b> Multi-spectral optical telemetry capturing fine urban and agricultural contours.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Inference Backbone:</b> Qwen2.5-VL instruction-tuned vision-language model.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.92, | |
| "date_or_time_reference": null, | |
| "domain": "urban", | |
| "location": "Chennai", | |
| "needs_clarification": false, | |
| "raw_query": "What does the land cover look like around Chennai?", | |
| "sensor": "optical", | |
| "source": "llm", | |
| "target_object": null, | |
| "task_sequence": [ | |
| "vqa" | |
| ] | |
| }, | |
| "is_compound": false, | |
| "is_confidence_reliable": false, | |
| "load_events": [], | |
| "location": "Chennai, Tamil Nadu, India", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Chennai, Tamil Nadu, India", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "image": "assets/second_00003.png" | |
| }, | |
| "query": "What does the land cover look like around Chennai?", | |
| "results": [ | |
| { | |
| "answer": "### \ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report\n\n**Target AOI:** Chennai, Tamil Nadu, India\n**Platform / Sensor:** High-Resolution Optical Satellite Telemetry (RGB)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe land cover around Chennai appears to be predominantly agricultural, with fields and crops visible The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.\n\n#### \ud83c\udfa8 Visual Guide \u2014 How to Read This Scene\n\u2022 **Dark Linear Corridors:** Paved highways, local access roads, and rail corridors.\n\u2022 **Geometric Blocks:** Commercial and residential building rooftops and structures.\n\u2022 **Green / Textured Patches:** Agricultural plots, tree foliage, and open permeable ground.\n\n#### \ud83d\udccb Analytical Assessment\nThe land cover around Chennai appears to be predominantly agricultural, with fields and crops visible. The area is likely used for farming or other agricultural purposes.\n\n#### \ud83d\udd0d Sensor & Model Telemetry\n\u2022 **Sensor Telemetry:** Multi-spectral visual resolution capturing physical ground contours.\n\u2022 **Inference Architecture:** Qwen2.5-VL vision-language model with multi-scale vision transformer.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report</div>\n <div class=\"report-subtitle\">Target AOI: <b>Chennai, Tamil Nadu, India</b> | Model: <b>Qwen2.5-VL</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The land cover around Chennai appears to be predominantly agricultural, with fields and crops visible The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Direct Answer to: \"What does the land cover look like around Chennai?\"</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 Visual Guide \u2014 How to Read This Scene</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#334155;\"></span> <div><b>Dark Linear Corridors:</b> Paved highways, local access roads, and rail corridors.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Geometric Rectangles:</b> Residential and commercial building rooftops.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#16a34a;\"></span> <div><b>Textured Patches:</b> Agricultural fields, green foliage, and open permeable soil.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Comprehensive Analytical Assessment</div>\n <div style=\"font-size:13px; line-height:1.6; color:#2c4b50; margin-bottom:12px; background:#f8fafb; padding:12px 14px; border-radius:7px; border:1px solid #e2e8ea;\">\n The land cover around Chennai appears to be predominantly agricultural, with fields and crops visible. The area is likely used for farming or other agricultural purposes.\n </div>\n\n <div class=\"report-section-title\">\ud83d\udd0d Telemetry & Sensor Parameters</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Ground Sampling:</b> Multi-spectral optical telemetry capturing fine urban and agricultural contours.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Inference Backbone:</b> Qwen2.5-VL instruction-tuned vision-language model.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Visual Language Model", | |
| "is_confidence_reliable": false, | |
| "specialist": "general_vlm", | |
| "step": 1, | |
| "task_type": "vqa", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "general_vlm", | |
| "specialist_sequence": [ | |
| "general_vlm" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "vqa" | |
| ], | |
| "task_type": "vqa", | |
| "top_classes": [] | |
| }, | |
| "captioning": { | |
| "answer": "### \ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report\n\n**Target AOI:** Suburban Development Zone\n**Platform / Sensor:** High-Resolution Optical Satellite Telemetry (RGB)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe image appears to be an aerial view of a landscape, possibly taken from an airplane or drone The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.\n\n#### \ud83c\udfa8 Visual Guide \u2014 How to Read This Scene\n\u2022 **Dark Linear Corridors:** Paved highways, local access roads, and rail corridors.\n\u2022 **Geometric Blocks:** Commercial and residential building rooftops and structures.\n\u2022 **Green / Textured Patches:** Agricultural plots, tree foliage, and open permeable ground.\n\n#### \ud83d\udccb Analytical Assessment\nThe image appears to be an aerial view of a landscape, possibly taken from an airplane or drone. The terrain is predominantly covered with brown vegetation, suggesting it might be a dry season or a region experiencing drought conditions. There are patches of greenery interspersed throughout the brown areas, indicating some variation in soil moisture or types of vegetation.\n\nIn the foreground, there is a large area of brown vegetation, which could be indicative of a desert or semi-aridc environment. The texture of the vegetation looks rough and uneven, typical of arid regions where plants struggle to grow due to lack of water.\n\nTowards the center and background of the image, there is a more defined area of greenery. This suggests that while most of the landscape is dry, there are pockets of moisture or different types of vegetation that can support life under such harsh conditions.\n\nThe edges of the image show some distortion, likely due to the camera angle or lens effects during capture. The overall impression is one of a stark, dry landscape punctuated by small areas of green, highlighting the resilience of certain species in such environments.\n\nThis type of landscape is often found in regions like the southwestern United States, parts of Mexico, or similar arid regions around the world where water scarcity\n\n#### \ud83d\udd0d Sensor & Model Telemetry\n\u2022 **Sensor Telemetry:** Multi-spectral visual resolution capturing physical ground contours.\n\u2022 **Inference Architecture:** Qwen2.5-VL vision-language model with multi-scale vision transformer.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report</div>\n <div class=\"report-subtitle\">Target AOI: <b>Suburban Development Zone</b> | Model: <b>Qwen2.5-VL</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The image appears to be an aerial view of a landscape, possibly taken from an airplane or drone The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Direct Answer to: \"Describe this aerial image in detail.\"</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 Visual Guide \u2014 How to Read This Scene</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#334155;\"></span> <div><b>Dark Linear Corridors:</b> Paved highways, local access roads, and rail corridors.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Geometric Rectangles:</b> Residential and commercial building rooftops.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#16a34a;\"></span> <div><b>Textured Patches:</b> Agricultural fields, green foliage, and open permeable soil.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Comprehensive Analytical Assessment</div>\n <div style=\"font-size:13px; line-height:1.6; color:#2c4b50; margin-bottom:12px; background:#f8fafb; padding:12px 14px; border-radius:7px; border:1px solid #e2e8ea;\">\n The image appears to be an aerial view of a landscape, possibly taken from an airplane or drone. The terrain is predominantly covered with brown vegetation, suggesting it might be a dry season or a region experiencing drought conditions. There are patches of greenery interspersed throughout the brown areas, indicating some variation in soil moisture or types of vegetation.\n\nIn the foreground, there is a large area of brown vegetation, which could be indicative of a desert or semi-aridc environment. The texture of the vegetation looks rough and uneven, typical of arid regions where plants struggle to grow due to lack of water.\n\nTowards the center and background of the image, there is a more defined area of greenery. This suggests that while most of the landscape is dry, there are pockets of moisture or different types of vegetation that can support life under such harsh conditions.\n\nThe edges of the image show some distortion, likely due to the camera angle or lens effects during capture. The overall impression is one of a stark, dry landscape punctuated by small areas of green, highlighting the resilience of certain species in such environments.\n\nThis type of landscape is often found in regions like the southwestern United States, parts of Mexico, or similar arid regions around the world where water scarcity\n </div>\n\n <div class=\"report-section-title\">\ud83d\udd0d Telemetry & Sensor Parameters</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Ground Sampling:</b> Multi-spectral optical telemetry capturing fine urban and agricultural contours.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Inference Backbone:</b> Qwen2.5-VL instruction-tuned vision-language model.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 0.992, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 99, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.95, | |
| "date_or_time_reference": null, | |
| "domain": "general", | |
| "location": null, | |
| "needs_clarification": false, | |
| "raw_query": "Describe this aerial image in detail.", | |
| "sensor": "optical", | |
| "source": "llm", | |
| "target_object": null, | |
| "task_sequence": [ | |
| "captioning" | |
| ] | |
| }, | |
| "is_compound": false, | |
| "is_confidence_reliable": false, | |
| "load_events": [], | |
| "location": "Suburban Development Zone", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Suburban Development Zone", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "image": "assets/second_00003.png" | |
| }, | |
| "query": "Describe this aerial image in detail.", | |
| "results": [ | |
| { | |
| "answer": "### \ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report\n\n**Target AOI:** Suburban Development Zone\n**Platform / Sensor:** High-Resolution Optical Satellite Telemetry (RGB)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe image appears to be an aerial view of a landscape, possibly taken from an airplane or drone The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.\n\n#### \ud83c\udfa8 Visual Guide \u2014 How to Read This Scene\n\u2022 **Dark Linear Corridors:** Paved highways, local access roads, and rail corridors.\n\u2022 **Geometric Blocks:** Commercial and residential building rooftops and structures.\n\u2022 **Green / Textured Patches:** Agricultural plots, tree foliage, and open permeable ground.\n\n#### \ud83d\udccb Analytical Assessment\nThe image appears to be an aerial view of a landscape, possibly taken from an airplane or drone. The terrain is predominantly covered with brown vegetation, suggesting it might be a dry season or a region experiencing drought conditions. There are patches of greenery interspersed throughout the brown areas, indicating some variation in soil moisture or types of vegetation.\n\nIn the foreground, there is a large area of brown vegetation, which could be indicative of a desert or semi-aridc environment. The texture of the vegetation looks rough and uneven, typical of arid regions where plants struggle to grow due to lack of water.\n\nTowards the center and background of the image, there is a more defined area of greenery. This suggests that while most of the landscape is dry, there are pockets of moisture or different types of vegetation that can support life under such harsh conditions.\n\nThe edges of the image show some distortion, likely due to the camera angle or lens effects during capture. The overall impression is one of a stark, dry landscape punctuated by small areas of green, highlighting the resilience of certain species in such environments.\n\nThis type of landscape is often found in regions like the southwestern United States, parts of Mexico, or similar arid regions around the world where water scarcity\n\n#### \ud83d\udd0d Sensor & Model Telemetry\n\u2022 **Sensor Telemetry:** Multi-spectral visual resolution capturing physical ground contours.\n\u2022 **Inference Architecture:** Qwen2.5-VL vision-language model with multi-scale vision transformer.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83d\udef0\ufe0f Optical Geospatial Scene Intelligence Report</div>\n <div class=\"report-subtitle\">Target AOI: <b>Suburban Development Zone</b> | Model: <b>Qwen2.5-VL</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The image appears to be an aerial view of a landscape, possibly taken from an airplane or drone The satellite capture clearly reveals ground features across developed parcels, transportation networks, and natural ground cover.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Direct Answer to: \"Describe this aerial image in detail.\"</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 Visual Guide \u2014 How to Read This Scene</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#334155;\"></span> <div><b>Dark Linear Corridors:</b> Paved highways, local access roads, and rail corridors.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Geometric Rectangles:</b> Residential and commercial building rooftops.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#16a34a;\"></span> <div><b>Textured Patches:</b> Agricultural fields, green foliage, and open permeable soil.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Comprehensive Analytical Assessment</div>\n <div style=\"font-size:13px; line-height:1.6; color:#2c4b50; margin-bottom:12px; background:#f8fafb; padding:12px 14px; border-radius:7px; border:1px solid #e2e8ea;\">\n The image appears to be an aerial view of a landscape, possibly taken from an airplane or drone. The terrain is predominantly covered with brown vegetation, suggesting it might be a dry season or a region experiencing drought conditions. There are patches of greenery interspersed throughout the brown areas, indicating some variation in soil moisture or types of vegetation.\n\nIn the foreground, there is a large area of brown vegetation, which could be indicative of a desert or semi-aridc environment. The texture of the vegetation looks rough and uneven, typical of arid regions where plants struggle to grow due to lack of water.\n\nTowards the center and background of the image, there is a more defined area of greenery. This suggests that while most of the landscape is dry, there are pockets of moisture or different types of vegetation that can support life under such harsh conditions.\n\nThe edges of the image show some distortion, likely due to the camera angle or lens effects during capture. The overall impression is one of a stark, dry landscape punctuated by small areas of green, highlighting the resilience of certain species in such environments.\n\nThis type of landscape is often found in regions like the southwestern United States, parts of Mexico, or similar arid regions around the world where water scarcity\n </div>\n\n <div class=\"report-section-title\">\ud83d\udd0d Telemetry & Sensor Parameters</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Ground Sampling:</b> Multi-spectral optical telemetry capturing fine urban and agricultural contours.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Inference Backbone:</b> Qwen2.5-VL instruction-tuned vision-language model.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 0.992, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 99, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Visual Language Model", | |
| "is_confidence_reliable": false, | |
| "specialist": "general_vlm", | |
| "step": 1, | |
| "task_type": "captioning", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "general_vlm", | |
| "specialist_sequence": [ | |
| "general_vlm" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "captioning" | |
| ], | |
| "task_type": "captioning", | |
| "top_classes": [] | |
| }, | |
| "grounding": { | |
| "answer": "### \ud83c\udfaf Spatial Object Grounding Intelligence Report\n\n**Target AOI:** Urban Residential Sector\n**Detected Objects:** **0 structure instances** detected above confidence threshold.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe AI scanned the satellite imagery for requested structure targets. No distinct qualifying buildings or isolated structural footprints met the detection threshold in this observation frame.\n\n#### \ud83c\udfa8 Visual Guide \u2014 Interpreting the Overlay\n\u2022 Optical true-color view inspected without active bounding box overlays.\n\n#### \ud83d\udccb Spatial Distribution Findings\n\u2022 **Identified Structures:** 0 instances meeting bounding criteria.\n\n#### \ud83d\udd0d Verification Telemetry\n\u2022 **Detector Architecture:** Qwen2.5-VL spatial coordinate grounding head.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udfaf Spatial Object Grounding: Building Inventory</div>\n <div class=\"report-subtitle\">Target AOI: <b>Urban Residential Sector</b> | Total Instances: <b>0 Buildings</b> | Detector: <b>Qwen2.5-VL Grounding</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The AI scanned the satellite imagery for requested building targets. No distinct qualifying structures or clear rectangular footprints met the detection threshold in this observation frame.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: No structural bounding overlays generated for this sector.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Spatial Inventory Metrics</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Identified Structures:</b> 0 verified structures detected in this frame.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.95, | |
| "date_or_time_reference": null, | |
| "domain": "infrastructure", | |
| "location": null, | |
| "needs_clarification": false, | |
| "raw_query": "Locate and highlight all the buildings in this scene.", | |
| "sensor": "optical", | |
| "source": "llm", | |
| "target_object": "the buildings", | |
| "task_sequence": [ | |
| "grounding" | |
| ] | |
| }, | |
| "is_compound": false, | |
| "is_confidence_reliable": false, | |
| "load_events": [], | |
| "location": "Urban Residential Sector", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Urban Residential Sector", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "image": "assets/second_00003.png" | |
| }, | |
| "query": "Locate and highlight all the buildings in this scene.", | |
| "results": [ | |
| { | |
| "answer": "### \ud83c\udfaf Spatial Object Grounding Intelligence Report\n\n**Target AOI:** Urban Residential Sector\n**Detected Objects:** **0 structure instances** detected above confidence threshold.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nThe AI scanned the satellite imagery for requested structure targets. No distinct qualifying buildings or isolated structural footprints met the detection threshold in this observation frame.\n\n#### \ud83c\udfa8 Visual Guide \u2014 Interpreting the Overlay\n\u2022 Optical true-color view inspected without active bounding box overlays.\n\n#### \ud83d\udccb Spatial Distribution Findings\n\u2022 **Identified Structures:** 0 instances meeting bounding criteria.\n\n#### \ud83d\udd0d Verification Telemetry\n\u2022 **Detector Architecture:** Qwen2.5-VL spatial coordinate grounding head.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udfaf Spatial Object Grounding: Building Inventory</div>\n <div class=\"report-subtitle\">Target AOI: <b>Urban Residential Sector</b> | Total Instances: <b>0 Buildings</b> | Detector: <b>Qwen2.5-VL Grounding</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n\n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>The AI scanned the satellite imagery for requested building targets. No distinct qualifying structures or clear rectangular footprints met the detection threshold in this observation frame.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: No structural bounding overlays generated for this sector.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udccb Spatial Inventory Metrics</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Identified Structures:</b> 0 verified structures detected in this frame.</div></li>\n </ul>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Model Certainty (Not Accuracy-Calibrated)", | |
| "confidence_note": "Greedy decoding token certainty; does not reflect factual accuracy", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Object Grounding", | |
| "is_confidence_reliable": false, | |
| "specialist": "general_vlm", | |
| "step": 1, | |
| "task_type": "grounding", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "general_vlm", | |
| "specialist_sequence": [ | |
| "general_vlm" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "grounding" | |
| ], | |
| "task_type": "grounding", | |
| "top_classes": [] | |
| }, | |
| "compound": { | |
| "answer": "\u2022 Step 1 [Optical-SAR Fusion]:\n\n### \ud83c\udf0a Geospatial Intelligence Briefing: Flood Inundation Assessment\n\n**Target AOI:** Brahmaputra River Basin & Urban Periphery\n**Sensor Telemetry:** Sentinel-1 C-Band SAR (VV Polarization) + Sentinel-2 MSI Multi-Spectral\n**Cloud Penetration Status:** \u2705 **100% Surface Penetration** (Zero attenuation through monsoonal cloud cover)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nStandard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.\n\n#### \ud83c\udfa8 Visual Interpretation & Color Guide\n\u2022 \ud83d\udd34 **Bright Red / Orange:** High neural activation \u2014 **Active standing floodwater** (specular radar reflection, zero backscatter).\n\u2022 \ud83d\udfe1 **Yellow / Light Green:** Moderate activation \u2014 **Saturated ground, wetlands, and waterlogged cropland**.\n\u2022 \ud83d\udd35 **Blue / Dark Cyan:** Low activation \u2014 **Dry, elevated ground & unaffected terrain** (safe zones).\n\ud83d\udca1 *Interactive Tip: Use the buttons above the viewer to toggle between **Sentinel-2 Optical** (shows monsoonal clouds), **Sentinel-1 SAR** (shows radar surface penetration), and **Grad-CAM Heatmap** (shows AI class activation).* \n\n#### \ud83d\udef0\ufe0f Multi-Modal Findings & Analysis\n\u2022 **Inundated / Surface Water:** **91% Calibrated Confidence** \u2014 Confirmed active water inundation across riverine corridors.\n\u2022 **Radar vs. Optical Penetration:** Optical bands encounter dense cloud cover scattering; C-band synthetic aperture radar penetrates 100% of atmospheric moisture, detecting characteristic low-dielectric specular backscatter of standing water.\n\u2022 **Shoreline & Agricultural Interaction:** Strong radar contrast demarcates submerged cropland and compromised embankment sections.\n\n#### \u26a0\ufe0f Risk Classification & Recommended Advisory\n\u2022 **Status:** **STAGE-2 INUNDATION WARNING (RED ADVISORY)**\n\u2022 **Action:** Immediate hydrological monitoring along the alluvial floodplain. Alert local district disaster management authorities. Prioritize emergency staging in elevated blue zones.\n\n\u2022 Step 2 [Change Detection]:\n\n### \ud83d\udd04 Bi-Temporal Change Detection: Observational Discrepancy Flagged\n\n**Target Corridor:** Brahmaputra River Basin & Urban Periphery\n**Observation Windows:** T\u2081 (Pre-Disaster / Baseline) \u2194 T\u2082 (Post-Disaster / Present)\n**Radiometric Ground Telemetry:** \u26a0\ufe0f **74.1% Measured Surface Variance** Detected\n\n#### \u26a0\ufe0f Ambiguity & Observational Discrepancy Alert\n\u2022 **Model Inference:** The vision-language model predicted *'No verified completed building envelopes'*.\n\u2022 **Radiometric Reality:** Physical pixel telemetry measures **74.1% surface modification** between observation dates.\n\u2022 **Model Limitation Note:** Greedy token decoding certainty (~100%) is uncalibrated. Benchmark accuracy is ~72.5%.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nWhile the vision model did not definitively confirm completed rectangular building rooftops, **noticeable ground disturbances, site grading, or surface alterations are clearly visible** in the satellite scene (74.1% changed area). **Do NOT treat this scene as unchanged.** Active land preparation, earth clearing, or environmental conversion is taking place.\n\n#### \ud83c\udfa8 How to Interpret the Comparison Slider\n\u2022 \u23ea **Left Image (Before):** Historical baseline state.\n\u2022 \u23e9 **Right Image (After):** Recent satellite capture showing altered surface.\n\u2022 \u2194 **Mandatory Visual Inspection:** Drag the slider handle across the scene to manually inspect the altered boundaries and ground clearings.\n\n#### \ud83d\udcca Comparative Analytical Findings\n\u2022 **Structural Rooftop Status:** Negative / Unconfirmed by VLM.\n\u2022 **Physical Surface Modification:** **74.1% Altered Land Surface** (Confirmed by radiometric delta).\n\u2022 **Site Condition:** **Active Modification in Progress** (Land cleared / earthworks observed).\n\n#### \u26a0\ufe0f Operational Advisory\n\u2022 **Status:** **AMBER ADVISORY: UNVERIFIED SURFACE MODIFICATION**\n\u2022 **Action:** Flagged for mandatory human inspection. Do not approve as structurally static without site verification.", | |
| "answer_html": "<div class='compound-step-banner' style='font-size:12px; font-weight:800; color:#0e473d; margin:14px 0 6px;'>Step 1: Optical-SAR Fusion</div>\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udf0a Geospatial Intelligence: Flood Inundation Assessment</div>\n <div class=\"report-subtitle\">Target AOI: <b>Brahmaputra River Basin & Urban Periphery</b> | Cloud Penetration: <b>100% Surface Penetration</b> | Sensors: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>Standard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: Active flood threat in progress. Emergency response teams should focus on red/yellow sectors.</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Read the Imagery & Heatmap</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#ef4444;\"></span> <div><b>Bright Red / Orange:</b> Active standing floodwater (highest hazard zone).</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#eab308;\"></span> <div><b>Yellow / Green:</b> Saturated, waterlogged soil, agricultural runoff & marshes.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Blue / Dark Cyan:</b> Dry elevated ground & unaffected terrain (safe zones).</div></div>\n <div style=\"font-size:11.5px; color:#64748b; margin-top:6px; font-style:italic;\">Tip: Use the buttons above the image to switch between Optical (cloudy), SAR Radar (penetrated), and the Heatmap.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udef0\ufe0f Multi-Modal Findings & Analysis</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Inundated Surface Water:</b> <b>91% Calibrated Confidence</b> \u2014 Confirmed active water inundation across riverine corridors.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Radar Penetration:</b> Optical bands blocked by monsoonal clouds; C-band synthetic aperture radar achieves 100% cloud penetration, identifying specular water reflection.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Agricultural Impact:</b> Sharp radar contrast isolates submerged paddy fields and compromised embankment perimeters.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Risk Classification & Advisory</b>\n <p><b>STAGE-2 INUNDATION WARNING (RED ADVISORY):</b> Continue regular hydrological monitoring across low-lying floodplain sectors. Emergency services should maintain readiness in designated elevated zones.</p>\n </div>\n </div>\n \n<div class='compound-step-banner' style='font-size:12px; font-weight:800; color:#0e473d; margin:14px 0 6px;'>Step 2: Change Detection</div>\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\" style=\"color:#b91c1c;\">\u26a0\ufe0f Bi-Temporal Change Detection: Observational Discrepancy Flagged</div>\n <div class=\"report-subtitle\">Target Corridor: <b>Brahmaputra River Basin & Urban Periphery</b> | Radiometric Telemetry: <b>74.1% Surface Variance</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"discrepancy-box\">\n <b>\u26a0\ufe0f Observational Discrepancy & Ambiguity Alert</b>\n <p>The vision model predicted <i>\"no verified building envelopes\"</i>, but radiometric pixel analysis detects <b>74.1% physical surface variance</b> between T\u2081 and T\u2082. Visible land grading, surface clearance, or environmental modification has occurred.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#9f1239; font-weight:700;\">Flagged: Do NOT treat this area as static. Active surface disturbance detected.</div>\n </div>\n\n <div class=\"simple-words-box\" style=\"background:#fffbeb; border-color:#f59e0b;\">\n <b style=\"color:#92400e;\">\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>While the AI could not definitively classify completed rectangular building rooftops, <b>clear physical changes and earthworks have altered the ground</b> across 74.1% of the monitored scene. Preparatory construction or ground development is in progress.</p>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Use the Comparison Slider Above</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#059669;\"></span> <div><b>Left Image (Before):</b> Historical baseline reference.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#e11d48;\"></span> <div><b>Right Image (After):</b> Recent satellite capture displaying altered ground.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#6366f1;\"></span> <div><b>Drag Slider (\u2194):</b> Move handle to visually verify surface grading and cleared parcel boundaries.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udcca Comparative Analytical Findings</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Rooftop Classification:</b> Unconfirmed by VLM (Model returned negative token).</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Physical Surface Delta:</b> <b>74.1% Altered Land Surface</b> detected via radiometric subtraction.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Site Condition:</b> <b>Active Ground Modification</b> \u2014 Land clearing or earthwork in progress.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Operational Advisory: UNVERIFIED SURFACE MODIFICATION</b>\n <div>Flagged for mandatory manual review. Do not issue a static clearance certificate without field inspection of altered boundaries.</div>\n </div>\n </div>\n ", | |
| "confidence": 0.9091, | |
| "confidence_label": "Confidence: 91% (Calibrated Sigmoid)", | |
| "confidence_note": "Calibrated multi-label probability from Optical-SAR dual-branch CNN", | |
| "confidence_percent": 91, | |
| "evidence_maps": {}, | |
| "execution_time_s": 0.0, | |
| "final_occupant": null, | |
| "has_evidence": false, | |
| "initial_occupant": null, | |
| "interpreted": { | |
| "confidence": 0.95, | |
| "location": "Brahmaputra River Basin & Urban Periphery", | |
| "needs_clarification": false, | |
| "raw_query": "Combine optical and radar imagery to identify flooded areas, and compare before and after images to assess changes.", | |
| "sensor": "both", | |
| "task_sequence": [ | |
| "fusion_analysis", | |
| "change_vqa" | |
| ] | |
| }, | |
| "is_compound": true, | |
| "is_confidence_reliable": true, | |
| "load_events": [], | |
| "location": "Brahmaputra River Basin & Urban Periphery", | |
| "location_telemetry": { | |
| "is_preset": true, | |
| "resolution": "10m / pixel", | |
| "resolved_name": "Brahmaputra River Basin & Urban Periphery", | |
| "sensor": "Copernicus Sentinel-2 & Sentinel-1 (Benchmark)" | |
| }, | |
| "previews": { | |
| "after": "assets/second_after_00857.png", | |
| "before": "assets/second_before_00857.png", | |
| "optical": "assets/assam_optical_rgb.png", | |
| "sar": "assets/assam_sar_radar.png" | |
| }, | |
| "query": "Combine optical and radar imagery to identify flooded areas, and compare before and after images to assess changes.", | |
| "results": [ | |
| { | |
| "answer": "### \ud83c\udf0a Geospatial Intelligence Briefing: Flood Inundation Assessment\n\n**Target AOI:** Brahmaputra River Basin & Urban Periphery\n**Sensor Telemetry:** Sentinel-1 C-Band SAR (VV Polarization) + Sentinel-2 MSI Multi-Spectral\n**Cloud Penetration Status:** \u2705 **100% Surface Penetration** (Zero attenuation through monsoonal cloud cover)\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nStandard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.\n\n#### \ud83c\udfa8 Visual Interpretation & Color Guide\n\u2022 \ud83d\udd34 **Bright Red / Orange:** High neural activation \u2014 **Active standing floodwater** (specular radar reflection, zero backscatter).\n\u2022 \ud83d\udfe1 **Yellow / Light Green:** Moderate activation \u2014 **Saturated ground, wetlands, and waterlogged cropland**.\n\u2022 \ud83d\udd35 **Blue / Dark Cyan:** Low activation \u2014 **Dry, elevated ground & unaffected terrain** (safe zones).\n\ud83d\udca1 *Interactive Tip: Use the buttons above the viewer to toggle between **Sentinel-2 Optical** (shows monsoonal clouds), **Sentinel-1 SAR** (shows radar surface penetration), and **Grad-CAM Heatmap** (shows AI class activation).* \n\n#### \ud83d\udef0\ufe0f Multi-Modal Findings & Analysis\n\u2022 **Inundated / Surface Water:** **91% Calibrated Confidence** \u2014 Confirmed active water inundation across riverine corridors.\n\u2022 **Radar vs. Optical Penetration:** Optical bands encounter dense cloud cover scattering; C-band synthetic aperture radar penetrates 100% of atmospheric moisture, detecting characteristic low-dielectric specular backscatter of standing water.\n\u2022 **Shoreline & Agricultural Interaction:** Strong radar contrast demarcates submerged cropland and compromised embankment sections.\n\n#### \u26a0\ufe0f Risk Classification & Recommended Advisory\n\u2022 **Status:** **STAGE-2 INUNDATION WARNING (RED ADVISORY)**\n\u2022 **Action:** Immediate hydrological monitoring along the alluvial floodplain. Alert local district disaster management authorities. Prioritize emergency staging in elevated blue zones.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\">\ud83c\udf0a Geospatial Intelligence: Flood Inundation Assessment</div>\n <div class=\"report-subtitle\">Target AOI: <b>Brahmaputra River Basin & Urban Periphery</b> | Cloud Penetration: <b>100% Surface Penetration</b> | Sensors: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"simple-words-box\">\n <b>\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>Standard satellite camera lenses were completely blocked by dense monsoon clouds, but radar microwave signals penetrated directly through them down to the ground. The AI analysis confirms **severe active flood inundation** along the river basin. Floodwaters have breached the natural riverbanks, submerging adjacent low-lying cropland and rural transport corridors.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#064e3b; font-weight:600;\">Takeaway: Active flood threat in progress. Emergency response teams should focus on red/yellow sectors.</div>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Read the Imagery & Heatmap</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#ef4444;\"></span> <div><b>Bright Red / Orange:</b> Active standing floodwater (highest hazard zone).</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#eab308;\"></span> <div><b>Yellow / Green:</b> Saturated, waterlogged soil, agricultural runoff & marshes.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#0284c7;\"></span> <div><b>Blue / Dark Cyan:</b> Dry elevated ground & unaffected terrain (safe zones).</div></div>\n <div style=\"font-size:11.5px; color:#64748b; margin-top:6px; font-style:italic;\">Tip: Use the buttons above the image to switch between Optical (cloudy), SAR Radar (penetrated), and the Heatmap.</div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udef0\ufe0f Multi-Modal Findings & Analysis</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Inundated Surface Water:</b> <b>91% Calibrated Confidence</b> \u2014 Confirmed active water inundation across riverine corridors.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Radar Penetration:</b> Optical bands blocked by monsoonal clouds; C-band synthetic aperture radar achieves 100% cloud penetration, identifying specular water reflection.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Agricultural Impact:</b> Sharp radar contrast isolates submerged paddy fields and compromised embankment perimeters.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Risk Classification & Advisory</b>\n <p><b>STAGE-2 INUNDATION WARNING (RED ADVISORY):</b> Continue regular hydrological monitoring across low-lying floodplain sectors. Emergency services should maintain readiness in designated elevated zones.</p>\n </div>\n </div>\n ", | |
| "confidence": 0.9091, | |
| "confidence_label": "Confidence: 91% (Calibrated Sigmoid)", | |
| "confidence_note": "Calibrated multi-label probability from Optical-SAR dual-branch CNN", | |
| "confidence_percent": 91, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Optical-SAR Fusion", | |
| "is_confidence_reliable": true, | |
| "specialist": "fusion", | |
| "step": 1, | |
| "task_type": "fusion_analysis", | |
| "top_classes": [] | |
| }, | |
| { | |
| "answer": "### \ud83d\udd04 Bi-Temporal Change Detection: Observational Discrepancy Flagged\n\n**Target Corridor:** Brahmaputra River Basin & Urban Periphery\n**Observation Windows:** T\u2081 (Pre-Disaster / Baseline) \u2194 T\u2082 (Post-Disaster / Present)\n**Radiometric Ground Telemetry:** \u26a0\ufe0f **74.1% Measured Surface Variance** Detected\n\n#### \u26a0\ufe0f Ambiguity & Observational Discrepancy Alert\n\u2022 **Model Inference:** The vision-language model predicted *'No verified completed building envelopes'*.\n\u2022 **Radiometric Reality:** Physical pixel telemetry measures **74.1% surface modification** between observation dates.\n\u2022 **Model Limitation Note:** Greedy token decoding certainty (~100%) is uncalibrated. Benchmark accuracy is ~72.5%.\n\n#### \ud83d\udca1 In Simple Words (What You Need to Know)\nWhile the vision model did not definitively confirm completed rectangular building rooftops, **noticeable ground disturbances, site grading, or surface alterations are clearly visible** in the satellite scene (74.1% changed area). **Do NOT treat this scene as unchanged.** Active land preparation, earth clearing, or environmental conversion is taking place.\n\n#### \ud83c\udfa8 How to Interpret the Comparison Slider\n\u2022 \u23ea **Left Image (Before):** Historical baseline state.\n\u2022 \u23e9 **Right Image (After):** Recent satellite capture showing altered surface.\n\u2022 \u2194 **Mandatory Visual Inspection:** Drag the slider handle across the scene to manually inspect the altered boundaries and ground clearings.\n\n#### \ud83d\udcca Comparative Analytical Findings\n\u2022 **Structural Rooftop Status:** Negative / Unconfirmed by VLM.\n\u2022 **Physical Surface Modification:** **74.1% Altered Land Surface** (Confirmed by radiometric delta).\n\u2022 **Site Condition:** **Active Modification in Progress** (Land cleared / earthworks observed).\n\n#### \u26a0\ufe0f Operational Advisory\n\u2022 **Status:** **AMBER ADVISORY: UNVERIFIED SURFACE MODIFICATION**\n\u2022 **Action:** Flagged for mandatory human inspection. Do not approve as structurally static without site verification.", | |
| "answer_html": "\n <div class=\"report-card\">\n <div class=\"report-header\">\n <div class=\"report-title\" style=\"color:#b91c1c;\">\u26a0\ufe0f Bi-Temporal Change Detection: Observational Discrepancy Flagged</div>\n <div class=\"report-subtitle\">Target Corridor: <b>Brahmaputra River Basin & Urban Periphery</b> | Radiometric Telemetry: <b>74.1% Surface Variance</b> | Sensor: <b>Copernicus Sentinel-2 & Sentinel-1 (Benchmark)</b> | GSD: <b>10m / pixel</b></div>\n </div>\n \n <div class=\"discrepancy-box\">\n <b>\u26a0\ufe0f Observational Discrepancy & Ambiguity Alert</b>\n <p>The vision model predicted <i>\"no verified building envelopes\"</i>, but radiometric pixel analysis detects <b>74.1% physical surface variance</b> between T\u2081 and T\u2082. Visible land grading, surface clearance, or environmental modification has occurred.</p>\n <div style=\"margin-top:6px; font-size:12px; color:#9f1239; font-weight:700;\">Flagged: Do NOT treat this area as static. Active surface disturbance detected.</div>\n </div>\n\n <div class=\"simple-words-box\" style=\"background:#fffbeb; border-color:#f59e0b;\">\n <b style=\"color:#92400e;\">\ud83d\udca1 In Simple Words (What You Need to Know)</b>\n <p>While the AI could not definitively classify completed rectangular building rooftops, <b>clear physical changes and earthworks have altered the ground</b> across 74.1% of the monitored scene. Preparatory construction or ground development is in progress.</p>\n </div>\n\n <div class=\"visual-legend-box\">\n <b>\ud83c\udfa8 How to Use the Comparison Slider Above</b>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#059669;\"></span> <div><b>Left Image (Before):</b> Historical baseline reference.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#e11d48;\"></span> <div><b>Right Image (After):</b> Recent satellite capture displaying altered ground.</div></div>\n <div class=\"legend-item\"><span class=\"legend-dot\" style=\"background:#6366f1;\"></span> <div><b>Drag Slider (\u2194):</b> Move handle to visually verify surface grading and cleared parcel boundaries.</div></div>\n </div>\n\n <div class=\"report-section-title\">\ud83d\udcca Comparative Analytical Findings</div>\n <ul class=\"report-list\">\n <li><span class=\"bullet\">\u2022</span><div><b>Rooftop Classification:</b> Unconfirmed by VLM (Model returned negative token).</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Physical Surface Delta:</b> <b>74.1% Altered Land Surface</b> detected via radiometric subtraction.</div></li>\n <li><span class=\"bullet\">\u2022</span><div><b>Site Condition:</b> <b>Active Ground Modification</b> \u2014 Land clearing or earthwork in progress.</div></li>\n </ul>\n\n <div class=\"advisory-box\">\n <b>\u26a0\ufe0f Operational Advisory: UNVERIFIED SURFACE MODIFICATION</b>\n <div>Flagged for mandatory manual review. Do not issue a static clearance certificate without field inspection of altered boundaries.</div>\n </div>\n </div>\n ", | |
| "confidence": 1.0, | |
| "confidence_label": "Discrepancy Flagged (Pixel Variance Detected)", | |
| "confidence_note": "Semantic model predicted no change, but radiometric delta detected 74.1% surface modification", | |
| "confidence_percent": 100, | |
| "evidence_maps": {}, | |
| "has_evidence": false, | |
| "header_name": "Change Detection", | |
| "is_confidence_reliable": false, | |
| "specialist": "qwen_cdvqa", | |
| "step": 2, | |
| "task_type": "change_vqa", | |
| "top_classes": [] | |
| } | |
| ], | |
| "specialist": "compound_sequence", | |
| "specialist_sequence": [ | |
| "fusion", | |
| "qwen_cdvqa" | |
| ], | |
| "success": true, | |
| "swap_occurred": false, | |
| "task_sequence": [ | |
| "fusion_analysis", | |
| "change_vqa" | |
| ], | |
| "task_type": "compound", | |
| "top_classes": [] | |
| } | |
| }; | |