image semantic segmentation ml backend example?

[imean131415]

Is there an example for image semantic segmentation ml backend? I am not sure, how to write the predict results to adapt to label studio. Thank you!

[makseq]

Please check this docs https://labelstud.io/tutorials/object-detector.html

And also this video might be helpful https://youtu.be/CyRe73VD4EE

[imean131415]

Thanks. But these examples are not for image segmentation, for instance using PSPNet or BiSeNet or HROCR as predicting model. Do you have any example for this? Thank you!

[KonstantinKorotaev]

Ho @imean131415 You can check this example. The main point is to use encode_rle method to convert the mask to RLE format that is used in Label Studio.

[LetSur]

Hi @KonstantinKorotaev I try to implement the referred example. How can I pass 'context' that contains interaction information (i.e., point or bounding box) to predict function? Thank you in advance! Thank you!

[KonstantinKorotaev]

Hi @LetSur You should use Webhook feature from Label Studio - it has all necessary data including context

[makseq]

Also please check this issue, it might be useful - https://github.com/heartexlabs/label-studio/issues/732#issuecomment-810731402

[makseq]

Closed as outdated, solved or no more actual issue.