Merge pull request #25 from drishyakarki/main

add robust support for new datasets classes loading and improve the usage documentation

README.md

@@ -38,6 +38,8 @@ Please refer to [install.md](docs/install.md) for installation and data preparat
 
 Please refer to [usage.md](docs/usage.md) for details of training and evaluation scripts.
 
+_Note: - We have added few documentation guides inside different parts of the library to make working with the library easier. Like this is an [instruction](configs/_base_/datasets/json/README.md) for class intialization inside the configs directory. So please try to follow them if you get issues while working with the library. :)_
+
 ## 🤝 Roadmap
 
 All the things that need to be done in the future is in [roadmap.md](docs/en/roadmap.md).

configs/_base_/datasets/json/README.md

@@ -0,0 +1,46 @@
+# Instructions regarding Classes Initialization for end to end methods like E2ESpot
+
+If you are working with new datasets, then you might need to change the classes list accordingly to adapt to that particular dataset labels.
+
+For instance, when you are working with OSL-Action Spotting Dataset "spotting-OSL", it has 17 classes so you will have the classes as 
+```python
+classes = [
+    "Ball out of play",
+    "Clearance",
+    "Corner",
+    "Direct free-kick",
+    "Foul",
+    "Goal",
+    "Indirect free-kick",
+    "Kick-off",
+    "Offside",
+    "Penalty",
+    "Red card",
+    "Shots off target",
+    "Shots on target",
+    "Substitution",
+    "Throw-in",
+    "Yellow card",
+    "Yellow->red card",
+]
+```
+in the ```video_dali.py``` and ```video_ocv.py``` files.
+
+But if you are working with, say [SN-Ball Action Spotting Dataset](https://huggingface.co/datasets/SoccerNet/SN-BAS-2025), then you need to make these changes in the ```video_dali.py``` and ```video_ocv.py``` files.
+
+```python
+[
+    "PASS",
+    "DRIVE",
+    "HEADER",
+    "HIGH PASS",
+    "OUT",
+    "CROSS",
+    "THROW IN",
+    "SHOT",
+    "BALL PLAYER BLOCK",
+    "PLAYER SUCCESSFUL TACKLE",
+    "FREE KICK",
+    "GOAL"
+]
+```

docs/usage.md

@@ -103,6 +103,24 @@ python tools/evaluate.py \
     dataset.test.metric=tight
 ```
 
+### If you don't see any results, you might need to specify the path to your results file
+
+```bash
+python tools/evaluate.py \
+    configs/e2espot/e2espot.py \
+    --cfg-options dataset.test.results=/outputs/e2e/rny008_gsm_150/results_spotting_test.recall.json.gz
+```
+
+#### So for a custom evaluation on test set, it might look something like this
+
+```bash
+python tools/evaluate.py \
+    configs/e2espot/e2espot.py \
+    --cfg-options dataset.test.path=/datasets/224p/test/annotations.json \
+    dataset.test.data_root=/datasets/224p/test \
+    dataset.test.results=outputs/e2e/rny008_gsm_150/results_spotting_test.recall.json.gz \
+    dataset.test.metric=tight
+```
 ## Visualization
 
 ```bash

oslactionspotting/core/utils/default_args.py

@@ -107,4 +107,4 @@ def get_default_args_train(model, train_loader, valid_loader, classes, trainer_t
             "train_loader": train_loader,
             "valid_loader": valid_loader,
             "classes": classes,
-        }
+        }