This repo is too complex

[flavienbwk]

Hi,

I try to get started with ML Backend in Label Studio. My personal appreciation is that the example provided by this repo is too complex :

• The docker and docker-compose configurations should be at the root of the repo (Docker is a standard and should probably be the one for LabelStudio)
• The repo should provide 1 or 2 simple examples (text and image classification for example)
• There are several requirements.txt, there should be only one (you could add comments in the requirements.txt for each example's dependencies)

Currently it's very time consuming to get into ML Backend even if it's a critical feature for labelers.

I hope Heartex can make it more simple in the future (I guess you are very busy).

Thanks, A really-liking-label-studio user

[davodogster]

I'm also interested in instance segmentation and semantic segmentation examples. The current MMdetection example only mentions Bounding box.

[niklub]

Hey, guys, do think it would be easier if we put every model example in a separate repo? With clear commands how to setup its own environment. Like this one

[makseq]

Closed as outdated and probably solved.