BUGFIX: Adjust IOU threshold to compute ConfusionMatrix in validation step of YOLO detection

[bavo96]

1. Check for Existing Contributions: I have explored existing PRs and haven't found a solution yet.
2. Elaborate Your Changes: When running model validation, for yolo detection, I found out that the ConfusionMatrix doesn't change the IOU threshold (iou_thres=0.45). Below is my code to run the validation process:

from ultralytics import YOLO

test_conf = {
            "data": "./data.yaml",
            "project": "./yolov8",
            "name": "my_project/test",
            "split": "test",
            "verbose": True,
            "plots": True,
            "exist_ok": True,
            "save_json": True,
            "conf": 0.3,
            "iou": 0.5,
        }

model_path = "./my_project/train/weights/best.pt"  # load a trained model (for validation)
model = YOLO(model_path)
metrics = model.val(**test_conf)
print(metrics.confusion_matrix.iou_thres)

When I traceback to the val.py file, I find this:

self.confusion_matrix = ConfusionMatrix(nc=self.nc, conf=self.args.conf) 

I have added the self.arg.iou to the iou_thres in the ConfusionMatrix. I think my PR will help users adjust the iou_thres using configuration.

3. Ultralytics Contributor License Agreement (CLA): I have read the CLA Document and I sign the CLA

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced accuracy in model validation by incorporating IoU threshold in Confusion Matrix calculation.

📊 Key Changes

• Modified the creation of the Confusion Matrix within the validation process to include an IoU (Intersection over Union) threshold as a parameter.

🎯 Purpose & Impact

• Purpose: To refine the precision of the model's validation step by considering the IoU threshold when computing the Confusion Matrix. This aligns the evaluation metrics more closely with real-world expectations of object detection models.
• Impact: Users can expect more accurate and meaningful insights from model validation. This change has the potential to improve how models are fine-tuned and evaluated, leading to better overall performance and reliability. 🚀

[github-actions]

CLA Assistant Lite bot All Contributors have signed the CLA. ✅

[github-actions]

👋 Hello @bavo96, thank you for submitting an Ultralytics YOLOv8 🚀 PR! To allow your work to be integrated as seamlessly as possible, we advise you to:

• ✅ Verify your PR is up-to-date with ultralytics/ultralytics main branch. If your PR is behind you can update your code by clicking the 'Update branch' button or by running git pull and git merge main locally.
• ✅ Verify all YOLOv8 Continuous Integration (CI) checks are passing.
• ✅ Update YOLOv8 Docs for any new or updated features.
• ✅ Reduce changes to the absolute minimum required for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." — Bruce Lee

See our Contributing Guide for details and let us know if you have any questions!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 75.12%. Comparing base (c47eec8) to head (db70b44).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #10254      +/-   ##
==========================================
- Coverage   78.95%   75.12%   -3.84%     
==========================================
  Files         121      121              
  Lines       15351    15351              
==========================================
- Hits        12120    11532     -588     
- Misses       3231     3819     +588     

Flag	Coverage Δ
Benchmarks	35.79% <100.00%> (-0.24%)	⬇️
GPU	37.94% <100.00%> (+0.01%)	⬆️
Tests	70.54% <100.00%> (-4.15%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bavo96]

I have read the CLA Document and I sign the CLA

[bavo96]

Can you do a quick check on this @glenn-jocher?

[glenn-jocher]

@bavo96 absolutely! I'll look into this right away and circle back with any findings or necessary actions. Thanks for bringing this to my attention! 😊 If any specific details or examples come up in the meantime that could help investigate, please feel free to share!

[bavo96]

Many thanks for your help @glenn-jocher I'll recheck the code for any new issues while you prepare for your next action.

[glenn-jocher]

You're welcome! 😊 Glad I could assist. Please do keep me updated if you come across anything else; happy to help further. Enjoy coding and good luck with your project!