Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+trained_models_streamvid/sva_01/best_weights.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text
+trained_models_streamvid/sva_02/best_weights.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text
+trained_models_streamvid/sva_03/best_weights.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text
+trained_models_streamvid/sva_04/best_weights.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text

trained_models_streamvid/a3_local_eval.json

@@ -0,0 +1,27 @@
+{
+  "id": "a3_local",
+  "test_accuracy": 0.8121693134307861,
+  "auc": 0.8850816046583242,
+  "metrics": {
+    "Fight": {
+      "precision": 0.8138297872340425,
+      "recall": 0.8095238095238095,
+      "f1": 0.8116710875331565
+    },
+    "No_Fight": {
+      "precision": 0.8105263157894737,
+      "recall": 0.8148148148148148,
+      "f1": 0.812664907651715
+    }
+  },
+  "streaming": {
+    "causal_accuracy": 0.9666666666666667,
+    "per_window_latency_ms": 1402.698891965944,
+    "p99_window_latency_ms": 1755.9232403410717,
+    "streaming_stability": 0.18120379745960236,
+    "vram_mb": 178.777344,
+    "max_streams": 126,
+    "ttd_windows_mean": null,
+    "n_clips": 30
+  }
+}

trained_models_streamvid/a3_streaming_eval.json

@@ -0,0 +1,11 @@
+{
+  "method": "sliding_window_clip (12fr step15, pure A3 HF)",
+  "causal_accuracy": 0.9,
+  "per_window_latency_ms": 1175.68588469973,
+  "p99_window_latency_ms": 1781.6566092828004,
+  "streaming_stability": 0.9748132824897766,
+  "vram_mb": 178.765824,
+  "max_streams": 126,
+  "ttd_windows_mean": null,
+  "n_clips": 30
+}

trained_models_streamvid/evaluation_metrics/sva_01_streamvid/classification_metrics.txt

@@ -0,0 +1,7 @@
+Performance Metrics (StreamViD-A):
+Label                        Precision        Recall      F1-Score       AUC-ROC
+----------------------------------------------------------------------
+Fight                           0.8073        0.8201        0.8136        0.8974
+No_Fight                        0.8172        0.8042        0.8107        0.8974
+Test Accuracy: 0.8122 | Val(best): 0.8148
+Latency: 1488.8 ms/clip | VRAM: 181 MB/stream | Max streams: 124

trained_models_streamvid/evaluation_metrics/sva_02_streamvid/classification_metrics.txt

@@ -0,0 +1,7 @@
+Performance Metrics (StreamViD-A):
+Label                        Precision        Recall      F1-Score       AUC-ROC
+----------------------------------------------------------------------
+Fight                           0.7783        0.8360        0.8061        0.8863
+No_Fight                        0.8229        0.7619        0.7912        0.8863
+Test Accuracy: 0.7989 | Val(best): 0.8413
+Latency: 1501.8 ms/clip | VRAM: 261 MB/stream | Max streams: 86

trained_models_streamvid/evaluation_metrics/sva_03_streamvid/classification_metrics.txt

@@ -0,0 +1,7 @@
+Performance Metrics (StreamViD-A):
+Label                        Precision        Recall      F1-Score       AUC-ROC
+----------------------------------------------------------------------
+Fight                           0.7810        0.8677        0.8221        0.8912
+No_Fight                        0.8512        0.7566        0.8011        0.8912
+Test Accuracy: 0.8122 | Val(best): 0.8333
+Latency: 1501.0 ms/clip | VRAM: 248 MB/stream | Max streams: 90

trained_models_streamvid/evaluation_metrics/sva_04_streamvid/classification_metrics.txt

@@ -0,0 +1,7 @@
+Performance Metrics (StreamViD-A):
+Label                        Precision        Recall      F1-Score       AUC-ROC
+----------------------------------------------------------------------
+Fight                           0.7979        0.8148        0.8063        0.8937
+No_Fight                        0.8108        0.7937        0.8021        0.8937
+Test Accuracy: 0.8042 | Val(best): 0.8228
+Latency: 1461.2 ms/clip | VRAM: 253 MB/stream | Max streams: 89

trained_models_streamvid/streamvid_all_experiments.csv

@@ -0,0 +1,5 @@
+Exp,DR,MotionDim,Heads,ValAcc,TestAcc,AUC,Latency_ms,VRAM_MB,MaxStreams,TrainParams,TrainTime_min
+sva_01,0.3,64,4,0.8148,0.8122,0.8974,1488.8,181,124,252130,91.0
+sva_02,0.2,64,4,0.8413,0.7989,0.8863,1501.8,261,86,252130,65.8
+sva_03,0.4,64,4,0.8333,0.8122,0.8912,1501.0,248,90,252130,83.1
+sva_04,0.3,128,8,0.8228,0.8042,0.8937,1461.2,253,89,322210,91.5

trained_models_streamvid/streamvid_comparison.csv

@@ -0,0 +1,6 @@
+Model,TestAccuracy,AUC,Latency_ms,VRAM_MB,MaxStreams
+sva_01,0.8122,0.8974,1488.8,181,124
+sva_02,0.7989,0.8863,1501.8,261,86
+sva_03,0.8122,0.8912,1501.0,248,90
+sva_04,0.8042,0.8937,1461.2,253,89
+movinet_a3,0.8466,0.0000,0.0,2500,9

trained_models_streamvid/sva_01/best_weights.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c5e59ccf030e88d6b37b3573342e94317f739331d253d2eaf45d09d0481bf2b
+size 36402053

trained_models_streamvid/sva_01/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "best_weights"
+all_model_checkpoint_paths: "best_weights"

trained_models_streamvid/sva_01/result.json

@@ -0,0 +1,102 @@
+{
+  "id": "sva_01",
+  "dr": 0.3,
+  "motion_dim": 64,
+  "heads": 4,
+  "bs": 8,
+  "lr": 0.001,
+  "test_accuracy": 0.8121693134307861,
+  "test_loss": 0.4251619279384613,
+  "best_val_accuracy": 0.8148148059844971,
+  "auc": 0.8973992889336805,
+  "metrics": {
+    "Fight": {
+      "precision": 0.8072916666666666,
+      "recall": 0.8201058201058201,
+      "f1": 0.8136482939632546
+    },
+    "No_Fight": {
+      "precision": 0.8172043010752689,
+      "recall": 0.8042328042328042,
+      "f1": 0.8106666666666665
+    }
+  },
+  "latency_ms": 1488.8128995895386,
+  "vram_mb": 181.445888,
+  "stream_max": 124,
+  "trainable_params": 252130,
+  "train_time_min": 90.95606636603674,
+  "history": {
+    "loss": [
+      0.4922955334186554,
+      0.37904539704322815,
+      0.3001636862754822,
+      0.2568304240703583,
+      0.19526740908622742,
+      0.17105503380298615,
+      0.156440868973732,
+      0.14436301589012146,
+      0.1440611332654953,
+      0.14189647138118744
+    ],
+    "accuracy": [
+      0.753333330154419,
+      0.8344444632530212,
+      0.8616666793823242,
+      0.8905555605888367,
+      0.9288889169692993,
+      0.9377777576446533,
+      0.9516666531562805,
+      0.9522222280502319,
+      0.9527778029441833,
+      0.9527778029441833
+    ],
+    "val_loss": [
+      0.5006689429283142,
+      0.43924063444137573,
+      0.4880889356136322,
+      0.5183476209640503,
+      0.41012468934059143,
+      0.42082610726356506,
+      0.42598623037338257,
+      0.42551207542419434,
+      0.42488211393356323,
+      0.42472216486930847
+    ],
+    "val_accuracy": [
+      0.7539682388305664,
+      0.8015872836112976,
+      0.7883597612380981,
+      0.7645502686500549,
+      0.8148148059844971,
+      0.8068783283233643,
+      0.8068783283233643,
+      0.8095238208770752,
+      0.8121693134307861,
+      0.8121693134307861
+    ],
+    "lr": [
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      1.0000000656873453e-05,
+      1.0000000656873453e-05,
+      1.0000001111620804e-06
+    ]
+  },
+  "streaming": {
+    "method": "sliding_window_clip (12fr step15, clip model verified)",
+    "causal_accuracy": 0.8666666666666667,
+    "per_window_latency_ms": 1196.7690986658756,
+    "p99_window_latency_ms": 1857.976896743931,
+    "streaming_stability": 0.23489409685134888,
+    "vram_mb": 189.59872,
+    "max_streams": 118,
+    "ttd_windows_mean": null,
+    "n_clips": 30
+  }
+}

trained_models_streamvid/sva_02/best_weights.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c21eea6d5b091f816945c2eee4333e453ec1944a37cd237e1bdc51e154e16aa7
+size 36402053

trained_models_streamvid/sva_02/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "best_weights"
+all_model_checkpoint_paths: "best_weights"

trained_models_streamvid/sva_02/result.json

@@ -0,0 +1,86 @@
+{
+  "id": "sva_02",
+  "dr": 0.2,
+  "motion_dim": 64,
+  "heads": 4,
+  "bs": 8,
+  "lr": 0.001,
+  "test_accuracy": 0.7989417910575867,
+  "test_loss": 0.4558309018611908,
+  "best_val_accuracy": 0.841269850730896,
+  "auc": 0.8862853783488703,
+  "metrics": {
+    "Fight": {
+      "precision": 0.7783251231527094,
+      "recall": 0.8359788359788359,
+      "f1": 0.8061224489795918
+    },
+    "No_Fight": {
+      "precision": 0.8228571428571428,
+      "recall": 0.7619047619047619,
+      "f1": 0.7912087912087912
+    }
+  },
+  "latency_ms": 1501.7825365066528,
+  "vram_mb": 261.049088,
+  "stream_max": 86,
+  "trainable_params": 252130,
+  "train_time_min": 65.7522145152092,
+  "history": {
+    "loss": [
+      0.48506832122802734,
+      0.35325169563293457,
+      0.29868435859680176,
+      0.2503489553928375,
+      0.1560179889202118,
+      0.14438208937644958,
+      0.13016852736473083
+    ],
+    "accuracy": [
+      0.757777750492096,
+      0.8472222089767456,
+      0.8733333349227905,
+      0.8911111354827881,
+      0.945555567741394,
+      0.9522222280502319,
+      0.9666666388511658
+    ],
+    "val_loss": [
+      0.41304412484169006,
+      0.39303022623062134,
+      0.4391767680644989,
+      0.39673367142677307,
+      0.40707921981811523,
+      0.4339171051979065,
+      0.42253944277763367
+    ],
+    "val_accuracy": [
+      0.8121693134307861,
+      0.8042327761650085,
+      0.7698412537574768,
+      0.841269850730896,
+      0.8148148059844971,
+      0.7751322984695435,
+      0.8015872836112976
+    ],
+    "lr": [
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      1.0000000656873453e-05
+    ]
+  },
+  "streaming": {
+    "causal_accuracy": 0.0,
+    "per_frame_latency_ms": 1184.3818389986786,
+    "p99_latency_ms": 1247.723960410803,
+    "streaming_stability": 0.0014266129583120346,
+    "vram_mb": 266.17344,
+    "max_streams": 84,
+    "ttd_frames_mean": null,
+    "n_clips": 30
+  }
+}

trained_models_streamvid/sva_03/best_weights.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10ce0542ba6c27845c0c0d1d24b9e06df4d40a41003203a6d4e34e6b09ea56e9
+size 36402053

trained_models_streamvid/sva_03/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "best_weights"
+all_model_checkpoint_paths: "best_weights"

trained_models_streamvid/sva_03/result.json

@@ -0,0 +1,86 @@
+{
+  "id": "sva_03",
+  "dr": 0.4,
+  "motion_dim": 64,
+  "heads": 4,
+  "bs": 8,
+  "lr": 0.001,
+  "test_accuracy": 0.8121693134307861,
+  "test_loss": 0.42684242129325867,
+  "best_val_accuracy": 0.8333333134651184,
+  "auc": 0.8911844573220234,
+  "metrics": {
+    "Fight": {
+      "precision": 0.780952380952381,
+      "recall": 0.8677248677248677,
+      "f1": 0.8220551378446114
+    },
+    "No_Fight": {
+      "precision": 0.8511904761904762,
+      "recall": 0.7566137566137566,
+      "f1": 0.8011204481792716
+    }
+  },
+  "latency_ms": 1501.0386943817139,
+  "vram_mb": 247.918592,
+  "stream_max": 90,
+  "trainable_params": 252130,
+  "train_time_min": 83.0935604373614,
+  "history": {
+    "loss": [
+      0.5022268891334534,
+      0.3799457848072052,
+      0.3163597881793976,
+      0.28626546263694763,
+      0.24991780519485474,
+      0.20802001655101776,
+      0.152094766497612,
+      0.14217142760753632,
+      0.12114695459604263
+    ],
+    "accuracy": [
+      0.7516666650772095,
+      0.8338888883590698,
+      0.8588888645172119,
+      0.875,
+      0.8916666507720947,
+      0.9127777814865112,
+      0.9377777576446533,
+      0.949999988079071,
+      0.9622222185134888
+    ],
+    "val_loss": [
+      0.40890344977378845,
+      0.3938610851764679,
+      0.41623687744140625,
+      0.3879639506340027,
+      0.405028373003006,
+      0.45406049489974976,
+      0.4411548972129822,
+      0.448883056640625,
+      0.4405975043773651
+    ],
+    "val_accuracy": [
+      0.8253968358039856,
+      0.8333333134651184,
+      0.8042327761650085,
+      0.8121693134307861,
+      0.8015872836112976,
+      0.7857142686843872,
+      0.7910053133964539,
+      0.7936508059501648,
+      0.8015872836112976
+    ],
+    "lr": [
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      1.0000000656873453e-05
+    ]
+  }
+}

trained_models_streamvid/sva_04/best_weights.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96477e8629d8c995e799563368d685436042cbe7d15a2305013b69718ac39f8d
+size 37243013

trained_models_streamvid/sva_04/checkpoint

@@ -0,0 +1,2 @@
+model_checkpoint_path: "best_weights"
+all_model_checkpoint_paths: "best_weights"

trained_models_streamvid/sva_04/result.json

@@ -0,0 +1,91 @@
+{
+  "id": "sva_04",
+  "dr": 0.3,
+  "motion_dim": 128,
+  "heads": 8,
+  "bs": 8,
+  "lr": 0.001,
+  "test_accuracy": 0.8042327761650085,
+  "test_loss": 0.43207594752311707,
+  "best_val_accuracy": 0.8227513432502747,
+  "auc": 0.8936759889140842,
+  "metrics": {
+    "Fight": {
+      "precision": 0.7979274611398963,
+      "recall": 0.8148148148148148,
+      "f1": 0.8062827225130891
+    },
+    "No_Fight": {
+      "precision": 0.8108108108108109,
+      "recall": 0.7936507936507936,
+      "f1": 0.8021390374331551
+    }
+  },
+  "latency_ms": 1461.1865520477295,
+  "vram_mb": 252.719872,
+  "stream_max": 89,
+  "trainable_params": 322210,
+  "train_time_min": 91.48943538268408,
+  "history": {
+    "loss": [
+      0.4934042692184448,
+      0.3712114095687866,
+      0.31744861602783203,
+      0.27573278546333313,
+      0.2035369575023651,
+      0.1771947592496872,
+      0.1674656867980957,
+      0.1538143754005432,
+      0.15466168522834778,
+      0.1501058042049408
+    ],
+    "accuracy": [
+      0.7622222304344177,
+      0.8311111330986023,
+      0.8622221946716309,
+      0.8777777552604675,
+      0.9238888621330261,
+      0.9316666722297668,
+      0.9377777576446533,
+      0.9494444727897644,
+      0.945555567741394,
+      0.9438889026641846
+    ],
+    "val_loss": [
+      0.41304031014442444,
+      0.40637221932411194,
+      0.45507538318634033,
+      0.4203333258628845,
+      0.40368151664733887,
+      0.41275754570961,
+      0.4150822162628174,
+      0.4196803867816925,
+      0.42488715052604675,
+      0.4245511293411255
+    ],
+    "val_accuracy": [
+      0.8015872836112976,
+      0.817460298538208,
+      0.7936508059501648,
+      0.8042327761650085,
+      0.8227513432502747,
+      0.8095238208770752,
+      0.8015872836112976,
+      0.8042327761650085,
+      0.8068783283233643,
+      0.8068783283233643
+    ],
+    "lr": [
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.0010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      0.00010000000474974513,
+      1.0000000656873453e-05,
+      1.0000000656873453e-05,
+      1.0000001111620804e-06
+    ]
+  }
+}

trained_models_streamvid/training_metrics/training_plots_sva_01/training_history.csv

@@ -0,0 +1,11 @@
+Epoch,TrainLoss,ValLoss,TrainAcc,ValAcc
+1,0.4922955334186554,0.5006689429283142,0.753333330154419,0.7539682388305664
+2,0.37904539704322815,0.43924063444137573,0.8344444632530212,0.8015872836112976
+3,0.3001636862754822,0.4880889356136322,0.8616666793823242,0.7883597612380981
+4,0.2568304240703583,0.5183476209640503,0.8905555605888367,0.7645502686500549
+5,0.19526740908622742,0.41012468934059143,0.9288889169692993,0.8148148059844971
+6,0.17105503380298615,0.42082610726356506,0.9377777576446533,0.8068783283233643
+7,0.156440868973732,0.42598623037338257,0.9516666531562805,0.8068783283233643
+8,0.14436301589012146,0.42551207542419434,0.9522222280502319,0.8095238208770752
+9,0.1440611332654953,0.42488211393356323,0.9527778029441833,0.8121693134307861
+10,0.14189647138118744,0.42472216486930847,0.9527778029441833,0.8121693134307861

trained_models_streamvid/training_metrics/training_plots_sva_02/training_history.csv

@@ -0,0 +1,8 @@
+Epoch,TrainLoss,ValLoss,TrainAcc,ValAcc
+1,0.48506832122802734,0.41304412484169006,0.757777750492096,0.8121693134307861
+2,0.35325169563293457,0.39303022623062134,0.8472222089767456,0.8042327761650085
+3,0.29868435859680176,0.4391767680644989,0.8733333349227905,0.7698412537574768
+4,0.2503489553928375,0.39673367142677307,0.8911111354827881,0.841269850730896
+5,0.1560179889202118,0.40707921981811523,0.945555567741394,0.8148148059844971
+6,0.14438208937644958,0.4339171051979065,0.9522222280502319,0.7751322984695435
+7,0.13016852736473083,0.42253944277763367,0.9666666388511658,0.8015872836112976

trained_models_streamvid/training_metrics/training_plots_sva_03/training_history.csv

@@ -0,0 +1,10 @@
+Epoch,TrainLoss,ValLoss,TrainAcc,ValAcc
+1,0.5022268891334534,0.40890344977378845,0.7516666650772095,0.8253968358039856
+2,0.3799457848072052,0.3938610851764679,0.8338888883590698,0.8333333134651184
+3,0.3163597881793976,0.41623687744140625,0.8588888645172119,0.8042327761650085
+4,0.28626546263694763,0.3879639506340027,0.875,0.8121693134307861
+5,0.24991780519485474,0.405028373003006,0.8916666507720947,0.8015872836112976
+6,0.20802001655101776,0.45406049489974976,0.9127777814865112,0.7857142686843872
+7,0.152094766497612,0.4411548972129822,0.9377777576446533,0.7910053133964539
+8,0.14217142760753632,0.448883056640625,0.949999988079071,0.7936508059501648
+9,0.12114695459604263,0.4405975043773651,0.9622222185134888,0.8015872836112976

trained_models_streamvid/training_metrics/training_plots_sva_04/training_history.csv

@@ -0,0 +1,11 @@
+Epoch,TrainLoss,ValLoss,TrainAcc,ValAcc
+1,0.4934042692184448,0.41304031014442444,0.7622222304344177,0.8015872836112976
+2,0.3712114095687866,0.40637221932411194,0.8311111330986023,0.817460298538208
+3,0.31744861602783203,0.45507538318634033,0.8622221946716309,0.7936508059501648
+4,0.27573278546333313,0.4203333258628845,0.8777777552604675,0.8042327761650085
+5,0.2035369575023651,0.40368151664733887,0.9238888621330261,0.8227513432502747
+6,0.1771947592496872,0.41275754570961,0.9316666722297668,0.8095238208770752
+7,0.1674656867980957,0.4150822162628174,0.9377777576446533,0.8015872836112976
+8,0.1538143754005432,0.4196803867816925,0.9494444727897644,0.8042327761650085
+9,0.15466168522834778,0.42488715052604675,0.945555567741394,0.8068783283233643
+10,0.1501058042049408,0.4245511293411255,0.9438889026641846,0.8068783283233643