Tensorflow lite model with fp16& int8 optimisations detections accuracy in android comparison with preview feature of HUB about hub HOT 2 CLOSED

@vipin-prabhakaran hello! Thanks for reaching out with your question. 🌟

When you're seeing a difference in accuracy between the model in the preview feature of Ultralytics HUB and the TensorFlow Lite model with FP16 & INT8 optimizations, it typically stems from the precision reduction inherent to these optimizations. Both FP16 (floating-point 16-bit) and INT8 (integer 8-bit) quantizations reduce the model size and computation requirements, making them highly beneficial for mobile devices like Android.

However, this benefit comes at a potential cost to accuracy and performance, as the model's weights are approximated to fit into a smaller, less precise numerical format. This is likely why you're observing a discrepancy in detection accuracy compared to the preview feature, which uses the full precision model.

To mitigate this issue, you might want to:

1. Experiment with only one type of optimization at a time to see which one impacts your accuracy the least.
2. Consider using more robust data for training, which can sometimes help in retaining accuracy post-optimization.

For specific instructions on how to optimize models while aiming to maintain high accuracy, please refer to our documentation at https://docs.ultralytics.com/hub.

Your journey to optimize models for mobile deployment is crucial, and although some trial and error might be inevitable, the gains in making your models portable and efficient are immense. Keep up the good work! 🚀

from hub.

👋 Hello there! We wanted to give you a friendly reminder that this issue has not had any recent activity and may be closed soon, but don't worry - you can always reopen it if needed. If you still have any questions or concerns, please feel free to let us know how we can help.

For additional resources and information, please see the links below:

• Docs: https://docs.ultralytics.com
• HUB: https://hub.ultralytics.com
• Community: https://community.ultralytics.com

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLO 🚀 and Vision AI ⭐

from hub.