Full-size proto code (optional)

[glenn-jocher]

Signed-off-by: Glenn Jocher [email protected]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced TensorFlow model upscaling capabilities in YOLOv5.

📊 Key Changes

• Optional full-size protos added via a commented out TFUpsample call.
• TFUpsample class updated to support scale factors that are multiples of 2, rather than only supporting a scale factor of 2.

🎯 Purpose & Impact

• 🎈 The optional code for full-size protos provides flexibility to experiment with different upscaling options during the model's prototyping phase, possibly enhancing the model's detection capabilities.
• 📈 The TFUpsample changes allow more versatile image scaling, which facilitates experimentation with different image resolutions and could lead to optimized performance and accuracy in object detection.

Overall, these changes promote flexible model refinement and potentially contribute to more finely-tuned object detection models for developers and users of the YOLOv5 TensorFlow implementation.