Two dimensional size=(h,w) AutoShape support

[glenn-jocher]

May resolve #9039

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to YOLOv5's model compatibility and input handling.

📊 Key Changes

• Added support for ClassificationModel in addition to DetectionModel.
• Modified AutoShape for input handling to accommodate non-square image sizing.
• Added a warning for cases when AutoShape is incompatible, particularly with the new ClassificationModel.
• Adjusted image size handling, allowing the size parameter to accept an integer and expand it to a square size tuple.
• Improved image resizing logic to maintain the aspect ratio.
• Made the image resizing process more consistent, regardless of whether it is executed on a PyTorch model.

🎯 Purpose & Impact

• These changes improve the flexibility of input data that can be passed to the models, including support for both square and non-square images.
• The distinction between ClassificationModel and DetectionModel allows users to work with different model types more effectively.
• The warning and input adjustments ensure users are aware of model compatibility issues, preventing potential errors.
• These changes should enhance user experience by providing clearer feedback and more robust input handling, leading to improved performance in various use cases.