Training results: result_4.3.0

.github/workflows/train.yml

@@ -134,7 +134,7 @@ jobs:
       - name: Download test dataset
         run: |
           uv run dvc get https://github.com/pyronear/pyro-dataset \
-            data/processed/yolo_test_corrected --rev v4.0.0-corrected \
+            data/processed/yolo_test --rev v4.3.0 \
             --out ./data/test/yolo_test
 
       - name: Run evaluation
@@ -150,7 +150,7 @@ jobs:
       - name: Download sequential test dataset
         run: |
           uv run dvc get https://github.com/pyronear/pyro-dataset \
-            data/processed/sequential_test --rev v4.0.0-corrected \
+            data/processed/sequential_test --rev v4.3.0 \
             --out ./data/test/sequential_test
 
       - name: Run sequential predictions on test
@@ -298,7 +298,7 @@ jobs:
           {opt_table}
 
           {baseline_note}
-          **Branch:** `{os.environ.get('RESULT_BRANCH', '')}` | **Test dataset:** pyronear/pyro-dataset @ v4.0.0-corrected""")
+          **Branch:** `{os.environ.get('RESULT_BRANCH', '')}` | **Test dataset:** pyronear/pyro-dataset @ v4.3.0""")
           PYEOF
           )
 

dvc.lock

@@ -61,9 +61,9 @@ stages:
     deps:
     - path: ./data/01_model_input/yolo_train_val
       hash: md5
-      md5: 409302377938ce2a82f5338f3145cbbc.dir
-      size: 1914167911
-      nfiles: 32201
+      md5: da8f531db6ac70886c2793c3a6f18f16.dir
+      size: 2321533653
+      nfiles: 37609
     - path: ./scripts/data/model_input/build.py
       hash: md5
       md5: 7607af10f59267cf9859cb1d372aff26
@@ -79,8 +79,8 @@ stages:
     outs:
     - path: ./data/02_models/yolo/best
       hash: md5
-      md5: 0e3617cd00b6a1985129110757bca1a0.dir
-      size: 44813862
+      md5: 43df3b52671b31dff13034cd7751eed0.dir
+      size: 45158669
       nfiles: 24
   build_manifest_yolo_best:
     cmd:
@@ -90,8 +90,8 @@ stages:
     deps:
     - path: ./data/02_models/yolo/best/
       hash: md5
-      md5: 0e3617cd00b6a1985129110757bca1a0.dir
-      size: 44813862
+      md5: 43df3b52671b31dff13034cd7751eed0.dir
+      size: 45158669
       nfiles: 24
     - path: ./scripts/model/yolo/build_manifest.py
       hash: md5
@@ -100,8 +100,8 @@ stages:
     outs:
     - path: ./data/03_reporting/yolo/best/
       hash: md5
-      md5: c01eb3ce79f23e0999d3220813c01881.dir
-      size: 13529
+      md5: 3c3d4c84c4d1bef3a95864873330834f.dir
+      size: 13509
       nfiles: 1
   export_yolo_best@onnx-cpu:
     cmd:
@@ -111,8 +111,8 @@ stages:
     deps:
     - path: ./data/02_models/yolo/best/
       hash: md5
-      md5: 0e3617cd00b6a1985129110757bca1a0.dir
-      size: 44813862
+      md5: 43df3b52671b31dff13034cd7751eed0.dir
+      size: 45158669
       nfiles: 24
     - path: ./scripts/model/yolo/export.py
       hash: md5
@@ -121,7 +121,7 @@ stages:
     outs:
     - path: ./data/02_models/yolo-export/best/onnx/cpu
       hash: md5
-      md5: 3e101e69ac72dc3109be7afd2b61aa4a.dir
+      md5: 3a5154579ba5f9beb866506a822fcdcb.dir
       size: 38532985
       nfiles: 1
   export_yolo_best@onnx-mps:
@@ -153,8 +153,8 @@ stages:
     deps:
     - path: ./data/02_models/yolo/best/
       hash: md5
-      md5: 0e3617cd00b6a1985129110757bca1a0.dir
-      size: 44813862
+      md5: 43df3b52671b31dff13034cd7751eed0.dir
+      size: 45158669
       nfiles: 24
     - path: ./scripts/model/yolo/export.py
       hash: md5
@@ -163,7 +163,7 @@ stages:
     outs:
     - path: ./data/02_models/yolo-export/best/ncnn/cpu
       hash: md5
-      md5: f6f1f2b316198f427a51b3b8335e7b7c.dir
+      md5: b2512c224a2ddc6fc9ccd2ba80130eb6.dir
       size: 38001906
       nfiles: 5
   export_yolo_best@ncnn-mps:
@@ -190,31 +190,30 @@ stages:
   fetch_model_input:
     cmd:
     - uv run dvc get https://github.com/pyronear/pyro-dataset 
-      data/processed/yolo_train_val_corrected --rev v4.0.0-corrected --out 
+      data/processed/yolo_train_val --rev v4.3.0 --out 
       ./data/01_model_input/yolo_train_val || uv run dvc get 
-      git@github.com:pyronear/pyro-dataset.git 
-      data/processed/yolo_train_val_corrected --rev v4.0.0-corrected --out 
-      ./data/01_model_input/yolo_train_val
+      git@github.com:pyronear/pyro-dataset.git data/processed/yolo_train_val 
+      --rev v4.3.0 --out ./data/01_model_input/yolo_train_val
     outs:
     - path: ./data/01_model_input/yolo_train_val
       hash: md5
-      md5: 409302377938ce2a82f5338f3145cbbc.dir
-      size: 1914167911
-      nfiles: 32201
+      md5: da8f531db6ac70886c2793c3a6f18f16.dir
+      size: 2321533653
+      nfiles: 37609
   fetch_sequential_val:
     cmd:
     - uv run dvc get https://github.com/pyronear/pyro-dataset 
-      data/processed/sequential_train_val/val --rev v4.0.0-corrected --out 
+      data/processed/sequential_train_val/val --rev v4.3.0 --out 
       ./data/01_model_input/sequential_train_val/val || uv run dvc get 
       git@github.com:pyronear/pyro-dataset.git 
-      data/processed/sequential_train_val/val --rev v4.0.0-corrected --out 
+      data/processed/sequential_train_val/val --rev v4.3.0 --out 
       ./data/01_model_input/sequential_train_val/val
     outs:
     - path: ./data/01_model_input/sequential_train_val/val
       hash: md5
-      md5: 652f9f7e73f297b782ed68e1b4844cce.dir
-      size: 671953151
-      nfiles: 11478
+      md5: df9cc34a9f322e39f85bda0d8e183463.dir
+      size: 855129379
+      nfiles: 14054
   evaluate_sequential_yolo_best:
     cmd:
     - uv run python ./scripts/model/yolo/evaluate_sequential.py --model-path 
@@ -252,12 +251,12 @@ stages:
     deps:
     - path: ./data/01_model_input/sequential_train_val/val
       hash: md5
-      md5: 652f9f7e73f297b782ed68e1b4844cce.dir
-      size: 671953151
-      nfiles: 11478
+      md5: df9cc34a9f322e39f85bda0d8e183463.dir
+      size: 855129379
+      nfiles: 14054
     - path: ./data/02_models/yolo/best/weights/best.pt
       hash: md5
-      md5: 550d1b91fb93e52cf233166b03405a10
+      md5: abae965ab4a3a761433ef043a1c2c8ee
       size: 19225626
     - path: predict_sequential.py
       hash: md5
@@ -266,9 +265,9 @@ stages:
     outs:
     - path: ./data/03_reporting/sequential/predictions_labels_val
       hash: md5
-      md5: 796a49a2802fa773158ca168a680f12c.dir
-      size: 109228
-      nfiles: 3340
+      md5: 6e7fa704398d621797344eb5f628ac2c.dir
+      size: 130190
+      nfiles: 4106
   optimize_sequential_val:
     cmd:
     - uv run python optimize_sequential.py --labels-dir 
@@ -278,19 +277,19 @@ stages:
     deps:
     - path: ./data/03_reporting/sequential/predictions_labels_val
       hash: md5
-      md5: 796a49a2802fa773158ca168a680f12c.dir
-      size: 109228
-      nfiles: 3340
+      md5: 6e7fa704398d621797344eb5f628ac2c.dir
+      size: 130190
+      nfiles: 4106
     - path: optimize_sequential.py
       hash: md5
-      md5: 115eb45048da93018f7ea86b5f954270
-      size: 8645
+      md5: a032d3a9877382e4ba2d74e5c529aaab
+      size: 9087
     outs:
     - path: ./data/03_reporting/sequential/grid_search_val.tsv
       hash: md5
-      md5: aa6f9da227c79d6ffa2903e7b9a3d18b
-      size: 1873
+      md5: 583fc929350796aa8a99a8855fa69a5a
+      size: 1926
     - path: ./data/03_reporting/sequential/grid_search_val_top20.tsv
       hash: md5
-      md5: dd8e0aa4c8bc678deb27acbdd911c065
-      size: 973
+      md5: d9f04f6a924a6a74ca8ea6833f7a8082
+      size: 988

dvc.yaml

@@ -3,11 +3,11 @@ stages:
     cmd:
     - >-
       uv run dvc get https://github.com/pyronear/pyro-dataset
-      data/processed/yolo_train_val_corrected --rev v4.0.0-corrected
+      data/processed/yolo_train_val --rev v4.3.0
       --out ./data/01_model_input/yolo_train_val
       ||
       uv run dvc get git@github.com:pyronear/pyro-dataset.git
-      data/processed/yolo_train_val_corrected --rev v4.0.0-corrected
+      data/processed/yolo_train_val --rev v4.3.0
       --out ./data/01_model_input/yolo_train_val
     outs:
     - ./data/01_model_input/yolo_train_val
@@ -53,11 +53,11 @@ stages:
     cmd:
     - >-
       uv run dvc get https://github.com/pyronear/pyro-dataset
-      data/processed/sequential_train_val/val --rev v4.0.0-corrected
+      data/processed/sequential_train_val/val --rev v4.3.0
       --out ./data/01_model_input/sequential_train_val/val
       ||
       uv run dvc get git@github.com:pyronear/pyro-dataset.git
-      data/processed/sequential_train_val/val --rev v4.0.0-corrected
+      data/processed/sequential_train_val/val --rev v4.3.0
       --out ./data/01_model_input/sequential_train_val/val
     outs:
     - ./data/01_model_input/sequential_train_val/val

results/eval_results.json

@@ -1,8 +1,8 @@
 {
-  "precision": 0.7801831566503715,
-  "recall": 0.7129963898916968,
-  "map50": 0.7478004027603852,
-  "map50_95": 0.45667696600893326,
+  "precision": 0.7669834327140582,
+  "recall": 0.7092668024439919,
+  "map50": 0.7292169261616344,
+  "map50_95": 0.4602580872227276,
   "model_dir": "data/02_models/yolo/best",
   "data": "data/test/yolo_test/data.yaml",
   "split": "test"

results/eval_sequential_results.json

@@ -1,12 +1,12 @@
 {
   "nb_consecutive_frames": 4,
   "conf_thresh": 0.25,
-  "tp": 115,
-  "fn": 4,
-  "fp": 13,
-  "tn": 97,
-  "recall": 0.9664,
-  "fpr": 0.1182,
-  "precision": 0.8984,
-  "f1": 0.9312
+  "tp": 152,
+  "fn": 3,
+  "fp": 8,
+  "tn": 134,
+  "recall": 0.9806,
+  "fpr": 0.0563,
+  "precision": 0.95,
+  "f1": 0.9651
 }