Update README.md

README.md

@@ -1,4 +1,145 @@
----
-license: cc-by-4.0
-pretty_name: 'SLR-NoteSense: A Dual-RGBC Sensor Dataset for Sri Lankan Banknote Recognition'
----
+{
+  "dataset_name": "SL-Banknote-RGBC",
+  "version": "1.0",
+  "title": "SL-Banknote-RGBC: Dual-Sensor RGBC Measurements of Sri Lankan Banknotes",
+  "description": "A dataset of repeated RGBC measurements collected from distinct physical Sri Lankan banknotes using two TCS34725 color sensors for embedded banknote-denomination recognition research.",
+  "license": {
+    "name": "Creative Commons Attribution 4.0 International",
+    "spdx_id": "CC-BY-4.0",
+    "url": "https://creativecommons.org/licenses/by/4.0/",
+    "attribution_required": true
+  },
+  "country": "Sri Lanka",
+  "currency": "Sri Lankan Rupee",
+  "denominations_lkr": [
+    20,
+    50,
+    100,
+    500,
+    1000,
+    5000
+  ],
+  "num_denominations": 6,
+  "num_physical_banknotes": 634,
+  "num_raw_measurements": 12668,
+  "num_stable_measurements": 12035,
+  "sensor_configuration": {
+    "sensor_model": "TCS34725",
+    "num_sensors": 2,
+    "channels": [
+      "red",
+      "green",
+      "blue",
+      "clear"
+    ],
+    "integration_time_ms": 50,
+    "gain": "4x",
+    "interface": "I2C",
+    "prototype_connections": {
+      "sensor_1": {
+        "sda_gpio": 17,
+        "scl_gpio": 18
+      },
+      "sensor_2": {
+        "sda_gpio": 8,
+        "scl_gpio": 9
+      }
+    }
+  },
+  "record_structure": {
+    "unit_of_raw_record": "One paired measurement from Sensor 1 and Sensor 2 for one physical banknote at one sample index.",
+    "fields": [
+      "label",
+      "note_id",
+      "scan_id",
+      "sample_index",
+      "s1_r",
+      "s1_g",
+      "s1_b",
+      "s1_clear",
+      "s1_nr",
+      "s1_ng",
+      "s1_nb",
+      "s2_r",
+      "s2_g",
+      "s2_b",
+      "s2_clear",
+      "s2_nr",
+      "s2_ng",
+      "s2_nb"
+    ],
+    "normalized_features": {
+      "s1_nr": "s1_r / s1_clear",
+      "s1_ng": "s1_g / s1_clear",
+      "s1_nb": "s1_b / s1_clear",
+      "s2_nr": "s2_r / s2_clear",
+      "s2_ng": "s2_g / s2_clear",
+      "s2_nb": "s2_b / s2_clear"
+    }
+  },
+  "physical_banknote_counts": {
+    "20": 102,
+    "50": 102,
+    "100": 106,
+    "500": 111,
+    "1000": 106,
+    "5000": 107
+  },
+  "quality_control": {
+    "settling_transient": "sample_index 0 was excluded from stable/model-ready data because it showed a consistent sensor-settling transient.",
+    "missing_label_repair": "60 missing labels were restored from denomination-specific source files: 20 rows for LKR 20, 20 rows for LKR 100, and 20 rows for LKR 5000.",
+    "short_sequences": "Some acquisition sequences contain fewer than the nominal 20 readings. These are retained when sufficient stable measurements are available.",
+    "grouping_rule": "Each physical banknote is uniquely identified using denomination + note_id because note_id values restart within each denomination."
+  },
+  "recommended_splits": {
+    "group_key": "denomination + note_id",
+    "guidance": "Keep all measurements from the same physical banknote within a single training, validation, or test partition to prevent leakage."
+  },
+  "technical_validation": {
+    "best_reference_model": "RBF Support Vector Machine",
+    "features": 8,
+    "rbf_svm_c": 5,
+    "rbf_svm_gamma": 0.06,
+    "group_aware_repeated_cv_accuracy": 0.9833,
+    "group_aware_repeated_cv_macro_f1": 0.9833,
+    "group_held_out_accuracy": 0.9685,
+    "group_held_out_macro_f1": 0.9696,
+    "python_cpp_prediction_agreement": "634/634"
+  },
+  "intended_use": [
+    "Sri Lankan banknote denomination recognition",
+    "embedded machine learning",
+    "color-sensor classification",
+    "sensor feature engineering",
+    "group-aware machine-learning evaluation",
+    "assistive-technology research"
+  ],
+  "not_intended_for": [
+    "counterfeit detection",
+    "financial authentication",
+    "banknote valuation",
+    "security-feature verification"
+  ],
+  "recommended_release_structure": {
+    "raw": "Original collected measurements.",
+    "cleaned": "Quality-controlled measurements with documented label repair.",
+    "derived": "Scan-level or banknote-level features used for model development.",
+    "metadata": "Dataset description, schema, license, and provenance.",
+    "code": "Preprocessing, validation, and baseline-model scripts."
+  },
+  "repository": {
+    "url": "TO_BE_ADDED",
+    "doi": "TO_BE_ADDED"
+  },
+  "citation": {
+    "dataset_paper_title": "TO_BE_ADDED",
+    "authors": [
+      "TO_BE_ADDED"
+    ],
+    "year": 2026
+  },
+  "contact": {
+    "name": "TO_BE_ADDED",
+    "email": "TO_BE_ADDED"
+  }
+}