Spaces:
Sleeping
Sleeping
Commit History
-
Sparkle cursor: pure CSS SVG data-URI cursor (8-ray cyan sparkle) — no JS, guaranteed visible
paulamartya25e39d025 -
Sparkle cursor: switch to head= param (real script tag in HTML head) — actually executes in browser
paulamartya253553db5 -
Sparkle cursor: rAF-animated particles, max z-index, 1.2s DOM delay, click burst
paulamartya25ba564d5 -
Fix cursor: restore real cursor visibility, sparkle glow + particles appear on top of it
paulamartya257001e4c -
Fix sparkle cursor: use js= param in gr.Blocks() instead of gr.HTML script tag (Gradio sanitizes script tags)
paulamartya252cc5cec -
Add sparkle cursor effect — cyan glow orb + trailing particles + click burst
paulamartya251e3099f -
UI Overhaul: premium dark TechSharthi-style design — animated mesh bg, glassmorphism cards, cyan glow tabs, hero header with live stats
paulamartya25c601589
-
Remove cloud_server/server.py — legacy demo server with plaintext access, never deployed, replaced by server/server_fhe.py on Render.com
paulamartya25b981f7f
-
Model Evaluation: update with TRUE scores from retrained model (Acc=91.19%, F1=93.40%, ROC-AUC=0.9656, FN=1)
paulamartya25c62d361 -
Cleanup: remove datasets+pyarrow deps (no longer needed with static eval tab)
paulamartya25515b83d -
Model Evaluation: replace broken dynamic tab with beautiful static metrics (real numbers from model_metrics.json)
paulamartya25b8290ab -
Fix: pin pyarrow==13.0.0 to fix PyExtensionType error with datasets==2.14.5
paulamartya25017ef7e -
Fix: evaluate_model auto-downloads test set from HF Hub (keremberke/chest-xray-classification)
paulamartya25e14d422
-
UI: remove secret key notification from Wake Server status card
paulamartya25cab375e -
README: remove Security Audit table
paulamartya254aace6e -
README: comprehensive docs with FHE architecture, security audit, full project docs
paulamartya2546ae083 -
Fix: restore HF Space YAML metadata in README, full docs in GITHUB_README.md
paulamartya25922ffc8 -
Update README: TRUE FHE architecture, deployment docs, security audit, full project docs
paulamartya25472ca9e -
Add: Wake Server button + background keep-alive thread (pings Render every 10 min)
paulamartya254ddf005 -
TRUE FHE: point to live Render.com server (securelens-1-m5kt.onrender.com)
paulamartya25369ffed -
Add server/ subfolder: TRUE FHE inference server for Render.com deployment
paulamartya25f857b91 -
Fix FHE server URL to paulamartya25/amartya_2509 space
paulamartya25b51c2e4 -
TRUE FHE: route inference to separate SecureLens-Server Space over HTTP
paulamartya254ce3714 -
Fix: pin pydantic==1.10.13 and fastapi==0.95.2 for Gradio 3.x compatibility
paulamartya254c6983e -
Fix HF Spaces deployment: enhanced 5-tab FHE app with best_model.pth via LFS
paulamartya2539b7ef4
-
Add TRUE FHE implementation with CKKS encryption
paulamartya2536dcf90 -
Remove model from git - will upload via web
paulamartya2570a5895 -
Remove dataset, use LFS for models
paulamartya257970ff7 -
Remove test data and specs, track models with LFS
paulamartya2582cc0b2 -
Fix: Use Gradio 3.50.2 with Blocks API and pinned dependencies
paulamartya255a47de1 -
Fix Gradio 2.9.4 API compatibility for HuggingFace Spaces
paulamartya25d7f0400 -
Fix port binding for Render deployment
paulamartya25cde310b -
Force Python 3.11.9 for Render compatibility
paulamartya2525bda26 -
Fix dependencies for Render deployment - use compatible Python 3.11
paulamartya254d766de -
Add Render deployment configuration
paulamartya2555a6d9c
-
Fix NoneType error - add lazy loading
paulamartya252de4c27
-
Use Git LFS for large model files
paulamartya255895cf2 -
Deploy to Hugging Face - exclude test dataset
paulamartya25361a48f -
Clean up project structure - remove unnecessary files
paulamartya25Removed: - Old app versions (app.py, app_gradio.py, app_gradio_clean.py) - Backup files (*.backup, *.attack_backup) - Test scripts (test_*.py, demo_*.py) - Fix scripts (fix_*.py, diagnose_*.py, inspect_*.py) - Temporary files (temp_metrics.png, securelens_visual_demo.png) - Personal files (Photos/, WhatsApp images, PPT) - Old documentation (FHE_FIXES_APPLIED.md, GRADCAM_FIX_SUMMARY.md, etc.) - Unused directories (client/, tests/, logs/, .kiro/, venv/) Kept: - Core app (app_gradio_enhanced.py) - Crypto layer (crypto_layer/) - Server modules (cloud_server/) - Trained model (best_model.pth, linear_weights.json) - Test dataset (data/chest_xray/test/) - Configuration (requirements.txt, .gitignore) - Essential docs (README.md, LICENSE, PROJECT_OVERVIEW.md) Project now contains only essential production files.4cf8019
-
Add comprehensive model evaluation metrics to Gradio interface
Your NameFeatures added: ✅ F1 Score, Precision, Recall, Accuracy metrics ✅ ROC-AUC curve visualization with matplotlib ✅ Confusion Matrix heatmap with seaborn ✅ New 'Model Evaluation' tab (5th tab) in Gradio UI ✅ TestDataset class for efficient batch evaluation ✅ Evaluates entire test set (624 images) ✅ Detailed metrics: TP, TN, FP, FN breakdown ✅ Beautiful styled HTML output with color-coded metrics Technical improvements: - Integrated sklearn metrics (accuracy_score, precision_score, recall_score, f1_score) - Added matplotlib & seaborn for professional visualizations - ROC curve with AUC score calculation - Confusion matrix with proper class labels - Proper error handling with traceback display UI fixes: - Fixed Gradio 6.0 CSS deprecation warning (moved to launch()) - Changed default port from 7860 to 7861 to avoid conflicts - Added temp_metrics.png to .gitignore Testing: - All 5 tabs functional: FHE Classification, Attack Demo, Comparison, GradCAM, Model Evaluation - Evaluation runs on fu…8225016 -
Add comprehensive model evaluation metrics to Gradio interface
Your NameFeatures added: - F1 Score, Precision, Recall, Accuracy metrics - ROC-AUC curve visualization - Confusion Matrix heatmap - New 'Model Evaluation' tab in Gradio UI - sklearn, matplotlib, seaborn integration - TestDataset class for batch evaluation - Evaluates entire test set with detailed metrics display Fixed: - Gradio 6.0 CSS parameter moved to launch() - Changed default port to 7861 to avoid conflicts141641b -
Replace nuclear attacks with realistic adversarial attacks
Your Name- Noise: Reduced from 500x to 15x (subtle sensor noise) - Brightness: Gamma correction instead of 10x saturation - Blur: Kernel max 9 instead of 151 (motion blur) - Contrast: 40% reduction instead of 99% blackout - NEW: Adversarial attack with sinusoidal pattern (JPEG-like) - Combined: Subtle mix of noise + blur + contrast - Attacks now look realistic while still flipping diagnosis - Demonstrates real-world adversarial ML threatsbcccfb7 -
Add nuclear-level attacks for demonstration
Your Name- Noise: 500x multiplier + salt & pepper destruction - Brightness: 10x multiplier for complete saturation - Blur: Kernel size up to 151 for feature obliteration - Contrast: 0.01-0.11 range for near-blackout - NEW: Combined attack (300x noise + blur + brightness) - Attacks now GUARANTEED to flip diagnosis above 60% intensity - Perfectly demonstrates FHE protection effectiveness0e18146 -
Remove PNG file for HuggingFace deployment
Your Name0bc68a5 -
Fixed the issues
Your Name9a940be
-
Fix CKKS engine implementation
Your Name0776a0f
-
Update README.md
c5a206d
-
Update UI theme and model evaluation outputs
Your Name80c6039
-
Add Hugging Face Space configuration
Your Nameff5e3f8 -
Configure Hugging Face deployment
Your Name7454830