Upload config.json with huggingface_hub

config.json

@@ -1,146 +1,208 @@
 {
-  "version": "2.0",
-  "description": "AI Content Detection - Stacking Meta-Classifier Weights",
-  "models": {
+  "version": "v5.1",
+  "description": "AI Content Detection meta-classifier weights (stacking ensemble)",
+  "modalities": [
+    "image",
+    "audio",
+    "text",
+    "video"
+  ],
+  "benchmark": {
     "image": {
-      "classifiers": [
-        "NYUAD-ComNets/NYUAD_AI-generated_images_detector",
-        "Organika/sdxl-detector",
-        "umm-maybe/AI-image-detector",
-        "dima806/ai_vs_real_image_detection"
-      ],
-      "meta_classifier": {
-        "scaler_mean": [
-          0.6074603661000239,
-          0.686688522024317,
-          0.3849442189093679,
-          0.769625436037109,
-          -2.7338574212581257,
-          2.353111819316856e-05
-        ],
-        "scaler_scale": [
-          0.4283660795399362,
-          0.42213362151536316,
-          0.26207683401251386,
-          0.24410297413117948,
-          0.30930081066524256,
-          3.9731677371366066e-05
-        ],
-        "lr_coefficients": [
-          1.7135966231796458,
-          1.2825863926730567,
-          -0.29594309495805504,
-          1.300517880868273,
-          2.5003933635884863,
-          0.5796105924507469
-        ],
-        "lr_intercept": 7.351305864946982,
-        "feature_names": [
-          "nyuad_score",
-          "sdxl_score",
-          "ai_image_score",
-          "dima806_score",
-          "fft_slope",
-          "fft_hf_ratio"
-        ],
-        "training_samples": 570,
-        "accuracy": 0.991
-      }
+      "accuracy": 0.98,
+      "tpr": 1.0,
+      "tnr": 0.965,
+      "f1": 0.978,
+      "samples": 204
     },
     "audio": {
-      "classifiers": [
-        "DavidCombei/wav2vec2-xls-r-1b-DeepFake-AI4TRUST",
-        "Gustking/wav2vec2-large-xlsr-deepfake-audio-classification"
-      ],
-      "meta_classifier": {
-        "scaler_mean": [
-          0.6209166406230846,
-          0.09860975675284862,
-          0.021718353962836168,
-          1924.8187628110218,
-          1258.8948959347454,
-          0.15988650152081954,
-          6620.699833048576
-        ],
-        "scaler_scale": [
-          0.4712932092514088,
-          0.03994814349451387,
-          0.007767299011189912,
-          328.93736914891036,
-          327.4988847517819,
-          0.047731468459419314,
-          266.2532958839088
-        ],
-        "lr_coefficients": [
-          1.809447067482119,
-          0.6736835103826301,
-          0.42671477059530283,
-          0.4348569443039551,
-          0.5215924234667307,
-          -0.9464732912412551,
-          0.44753673298714475
-        ],
-        "lr_intercept": 1.9856021189388406,
-        "feature_names": [
-          "davidcombei_score",
-          "gustking_score",
-          "spectral_flatness",
-          "centroid_mean",
-          "centroid_std",
-          "zcr",
-          "rolloff_99"
-        ],
-        "training_samples": 300,
-        "accuracy": 0.82
-      }
+      "accuracy": 0.852,
+      "tpr": 0.85,
+      "tnr": 0.855,
+      "f1": 0.866,
+      "samples": 142
     },
     "text": {
-      "classifiers": [
-        "tiiuae/falcon-7b",
-        "tiiuae/falcon-7b-instruct",
-        "Hello-SimpleAI/chatgpt-detector-roberta"
-      ],
-      "meta_classifier": {
-        "scaler_mean": [
-          1.130450439453125,
-          0.12016203999519348,
-          -0.46647878605266724,
-          6.471777786621135,
-          0.5225984308983255,
-          0.35815911186395905,
-          4.67180587427633
-        ],
-        "scaler_scale": [
-          0.10663201835331575,
-          0.319080203489424,
-          0.1695241896637113,
-          0.3085286068821368,
-          0.10418367918547129,
-          0.11425552726076238,
-          0.4851605144906258
-        ],
-        "lr_coefficients": [
-          -1.7326994966116347,
-          -0.23487951489322348,
-          -1.1207961610247337,
-          0.07210147273481866,
-          1.197535850075283,
-          1.0243602811395163,
-          0.5908107658969592
-        ],
-        "lr_intercept": -0.40390869475088775,
-        "feature_names": [
-          "binoculars_score",
-          "roberta_ai_score",
-          "burstiness",
-          "entropy",
-          "ttr",
-          "hapax_ratio",
-          "avg_word_len"
-        ],
-        "training_samples": 80,
-        "accuracy": 0.975
-      }
+      "accuracy": 0.93,
+      "tpr": 0.917,
+      "tnr": 0.947,
+      "f1": 0.936,
+      "samples": 129
+    },
+    "video": {
+      "accuracy": 0.9,
+      "tpr": 0.833,
+      "tnr": 1.0,
+      "samples": 30
+    },
+    "overall": {
+      "accuracy": 0.928,
+      "samples": 475
     }
+  },
+  "image": {
+    "n_features": 8,
+    "feature_names": [
+      "NYUAD_ViT",
+      "ai_vs_real_ViT",
+      "DFDC_ViT",
+      "ai_image_det_ViT",
+      "fft_slope",
+      "hf_ratio",
+      "Bombek1_SigLIP2_DINOv2",
+      "SPAI_CVPR2025"
+    ],
+    "scaler_mean": [
+      0.46721075337286583,
+      0.4332848905084707,
+      0.34848470501282125,
+      0.7513610315914312,
+      -2.7428234702735845,
+      1.4757695660114816e-05,
+      0.47213903127932083,
+      0.5310949190042461
+    ],
+    "scaler_scale": [
+      0.4562829992667211,
+      0.4653274721438903,
+      0.2594560381028844,
+      0.2566914952700282,
+      0.31761878154208484,
+      1.745336794888413e-05,
+      0.4468171423032323,
+      0.4707389622737817
+    ],
+    "lr_coef": [
+      0.6488963010751596,
+      0.19470730198227582,
+      0.3669096091179738,
+      -1.1058065882150858,
+      -0.47635552888598026,
+      -0.015401252102331365,
+      2.5029078795863406,
+      1.237011726618108
+    ],
+    "lr_intercept": -0.7403570533419102,
+    "hf_models": [
+      "Nahrawy/AIorNot",
+      "haywoodsloan/ai-image-detector-deploy",
+      "prithivMLmods/Deep-Fake-Detector-Model",
+      "Organika/sdxl-detector",
+      "Bombek/BombDetector-v3-SigLIP2-DINOv2-LoRA"
+    ],
+    "notes": "Bombek1 is #1 feature (coef=+2.50). SPAI uses local repo."
+  },
+  "audio": {
+    "n_features": 9,
+    "feature_names": [
+      "DavidCombei_wav2vec2",
+      "Gustking_wav2vec2",
+      "mo_thecreator_deepfake",
+      "spectral_flatness",
+      "centroid_mean",
+      "centroid_std",
+      "zero_crossing_rate",
+      "spectral_rolloff",
+      "DF_Arena_1B"
+    ],
+    "scaler_mean": [
+      0.5667607612050348,
+      0.2773010993612484,
+      0.23310774392822925,
+      0.03141037016224877,
+      1807.2398348786571,
+      897.18004887457,
+      0.12301036345108962,
+      6620.40736210088,
+      0.5433762406366287
+    ],
+    "scaler_scale": [
+      0.48680867334512096,
+      0.29197482864644153,
+      0.4211570130989059,
+      0.024618810573647662,
+      459.40344999868597,
+      394.8528855416117,
+      0.046570088698838365,
+      829.6553459300637,
+      0.4155082795685684
+    ],
+    "lr_coef": [
+      0.7845433297452213,
+      -0.25601227158569434,
+      0.38715143588917217,
+      0.5305971113288093,
+      0.14191280089652655,
+      1.7648106776858394,
+      -1.6174243839603224,
+      -1.09787021389514,
+      1.092684667819162
+    ],
+    "lr_intercept": 0.39250921446958165,
+    "hf_models": [
+      "DavidCombei/wav2vec2-xls-r-1b-deepfake",
+      "Gustking/ai_voice_detect_xlsr",
+      "mo-thecreator/Deepfake-audio-detection",
+      "Aloukik21/DF_Arena_1B"
+    ]
+  },
+  "text": {
+    "n_features": 8,
+    "feature_names": [
+      "binoculars_score",
+      "roberta_chatgpt_score",
+      "perplexity_norm",
+      "avg_sentence_length",
+      "lexical_diversity",
+      "hapax_ratio",
+      "avg_word_length",
+      "fakespot_roberta_score"
+    ],
+    "scaler_mean": [
+      1.1353826005329457,
+      0.33250804246780497,
+      -0.48164806951384675,
+      5.916446148470062,
+      0.6490103211442594,
+      0.5124573713819743,
+      5.220866125485708,
+      0.6364287314816944
+    ],
+    "scaler_scale": [
+      0.19535976595611237,
+      0.45007809250809544,
+      0.21119484430166974,
+      1.1937958293169302,
+      0.19352867829552858,
+      0.21389850106439456,
+      1.2135677101079925,
+      0.43094435530407293
+    ],
+    "lr_coef": [
+      -0.6243579398646565,
+      0.389259232075374,
+      -0.5040499517552531,
+      -0.21291399657541557,
+      -0.08360375807827485,
+      -0.014109874794709326,
+      0.22446151217916235,
+      1.2266905154327146
+    ],
+    "lr_intercept": 0.1964292008569683,
+    "hf_models": [
+      "tiiuae/falcon-7b",
+      "tiiuae/falcon-7b-instruct",
+      "roberta-base-openai-detector",
+      "fakespot-ai/roberta-base-fakespot-detector"
+    ],
+    "notes": "Binoculars uses falcon-7b pair. fakespot is #1 feature (coef=+1.23)."
+  },
+  "video": {
+    "method": "frame_sampling + audio_extraction",
+    "frame_detector": "image meta-classifier on sampled frames",
+    "audio_detector": "audio meta-classifier on extracted audio (ffmpeg)",
+    "combining": "50/50 equal weight of video_prob and audio_prob",
+    "default_num_frames": 8,
+    "notes": "Video = image detection on frames + audio detection on extracted track"
   }
 }