Update/inplace ops

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to dataset preprocessing, image normalization, and logging optimizations in the YOLOv5 repository.

📊 Key Changes

• Modified bounding box conversion process in Objects365.yaml dataset script for better accuracy and clipping.
• Optimized image normalization operation in detect.py.
• Simplified confidence accumulation logic in WandB integration (wandb_utils.py).

🎯 Purpose & Impact

• Ensuring bounding boxes are normalized and clipped properly increases the quality of training data, leading to potentially more accurate object detection models. 🎨
• Streamlining the image division operation improves the performance of the detection script, reducing computational overhead. ⚡
• The accumulation change in WandB logging provides cleaner, more efficient code, and could marginally improve logging performance. ✨

Overall, these updates are expected to improve efficiency and accuracy of the models trained using YOLOv5, with better data handling and streamlined codebase contributing to a more seamless user experience. 🚀