panp4n

> You can try passing `retina_masks=True` It works, Thank you so much!  I need optimize my inference code, How can I implement retina_masks in my code.

> @panp4n you're welcome! To implement `retina_masks` in your code, you can set the parameter when calling the model's predict function, like this: `model.predict(source, retina_masks=True)`. I knew that, I want...

> The code used by `retina_masks` is here > > https://github.com/ultralytics/ultralytics/blob/8154a27e9b332dd2dac7a107003462f0e20f6179/ultralytics/utils/ops.py#L689 > > https://github.com/ultralytics/ultralytics/blob/8154a27e9b332dd2dac7a107003462f0e20f6179/ultralytics/models/yolo/segment/predict.py#L49-L50 Thank you! This helps me a lot.

> The code used by `retina_masks` is here > > https://github.com/ultralytics/ultralytics/blob/8154a27e9b332dd2dac7a107003462f0e20f6179/ultralytics/utils/ops.py#L689 > > https://github.com/ultralytics/ultralytics/blob/8154a27e9b332dd2dac7a107003462f0e20f6179/ultralytics/models/yolo/segment/predict.py#L49-L50 https://github.com/ultralytics/ultralytics/blob/8154a27e9b332dd2dac7a107003462f0e20f6179/ultralytics/utils/ops.py#L729 I can't reproduce this `align_corners=False` by using OpenCV C++ api . The example also uses...

The same model, why the mask of official predict script is better than deployment code? Do you have any recommendations?

It works with CLI `yolo predict model: F:\v8test\test11\weights\last.pt source: F:\ultralytics-main\datasets\test11\images\val retina_masks: True ` How can I use yaml config to predict?

8.0 and 8.1 version are fine. The yaml path is correct.

I Tried. The same error.

The same as newer `default.yaml`. and it works on 8.0.1.6. Is there any changes in the predict code? Train and export is fine with this CLI.

Path to an image is the same error. The path is correct.