Max Tkachenko comments

Results 437 comments of


                                            Max Tkachenko

How to make ml-backend for segmentation model

You are using PolygonLabels, it's not popular case unlike bboxes. Looks like you need to integrate something like UNet model (pixel masks) and then convert image masks to polygons. As...

Validation error from label-studio after docker-compose up -d of ml-backend

Have you set up your labeling config in Label Studio?

Validation error from label-studio after docker-compose up -d of ml-backend

Also it would be great if you could share ML logs.

Validation error from label-studio after docker-compose up -d of ml-backend

@fishfree Just to copy logs from your ML backend terminal.

Validation error from label-studio after docker-compose up -d of ml-backend

Closed as non-active.

raise KeyError(key) from None KeyError: 'config_file'

Try to explicitly specify config_file as an environment variable: ``` config_file=/path/to/config/faster_rcnn_r50_fpn_1x.py label-studio-ml init coco-detector config_file=/path/to/config/faster_rcnn_r50_fpn_1x.py label-studio-ml start coco-detector ```

semaphore_tracker: process died unexpectedly

Please provide more details. What model do you use?

Using Ctrl + click to select multiple regions initiate another region drawing

@Rafeethu Try to unselect label button beforehand. Also you can play with these two options: ![image](https://user-images.githubusercontent.com/501780/166835532-df3feeb4-6db8-4ab8-8e56-7d015ae9456c.png)

Could you add question answering example code?

We'll be happy to add this model if you contribute!

ENH: Add support for reST-style docstrings

@kernc @collerek Why don't you merge it? :slightly_frowning_face: