Cannot find clear documentation on mp.Image, is the expected input BGR or RGB?

[MatoFD]

Cannot find clear documentation on mp.Image, is the expected input BGR or RGB?

[kuaashish]

Hi @MatoFD,

You can use both mp_image and mp_image2. When providing the rgb_image to segmentation_result = segmenter.segment(rgb_image), make sure you pass either mp_image or mp_image2. OpenCV images must be converted to a MediaPipe image first for compatibility since they are not directly supported. As for accuracy-related queries, we currently don't have metrics available for measuring segmentation accuracy.

Among the mp_image and mp_image_2, you can use both but generally we would recommend to feed RGB image to mp.Image() since it has a better segmentation among similar color gradients, however we don't have concrete data to back this up.

Thank you!!

[MatoFD]

Thanks! It's good to have a recommended solution and it's working well for now

[kuaashish]

Hi @MatoFD,

Thank you for confirming, Can we now proceed to mark this matter as resolved and close the status?

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No