Compute TP, FP from Precision, Recall, and target count. Leave values in val.py.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced average precision computation in object detection metrics for clarity and reliability.

📊 Key Changes

• Added an eps parameter to prevent division by zero in calculations.
• Refactored to use the true label counts directly from unique class counts for better performance.
• Introduced explicit true and false positive count outputs to the precision-recall curve function.
• Updated associated functions to handle the new outputs on true and false positives.

🎯 Purpose & Impact

• 🎯 Purpose: To improve the numerical stability and accuracy of the object detection metrics.
• 🚀 Impact: Users will experience more reliable performance metrics, which are crucial for evaluating and improving object detection models.
• 🧑‍💻 Developers: Can easily interpret the impact of their models' precision and recall due to the clearer computation of true and false positives.
• 🛠️ Overall: This change facilitates a more robust and transparent evaluation of object detection systems.