https://github.com/matthiaskoenig/labelstudio
Setup of label studio
https://github.com/matthiaskoenig/labelstudio
Last synced: 4 months ago
JSON representation
Setup of label studio
- Host: GitHub
- URL: https://github.com/matthiaskoenig/labelstudio
- Owner: matthiaskoenig
- Created: 2024-01-25T14:23:50.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-06-24T09:19:43.000Z (12 months ago)
- Last Synced: 2025-01-04T20:24:35.622Z (5 months ago)
- Language: Python
- Size: 40 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# label-studio-tools
This repository provides an interface for label-studio to programatically
manage annotation tasks. Main focus is the annotation of histological images.
This uses the label-studio SDK to manage the annotation tasks.
https://labelstud.io/blog/5-tips-and-tricks-for-label-studio-s-api-and-sdk/
https://labelstud.io/guide/sdk
# Deployment
For server deployment information see: [./DEPLOY.md](./DEPLOY.md)
# Installation
## virtual environment
Create a virtual environment and install the dependencies:
```bash
mkvirtualenv labelstudio --python=python3.11
(labelstudio) pip install -r requirements.txt
```
## SDK url and key
Create an `.env` file and add the `API_URL` and `API_KEY`.
The `API_KEY` is available from the label-studio interface under user settings.
```
cp .env.template .env
```
## Test connection
To test the SDK connection use the `sdktest.py` script.
## Setup trainings database macrosteatosis
- update API key from labelstudio; https://annotatedb.com/user/account
- copy images & raw predictions to folder and set the folder in the `.env` file
- Upload images to server; -> File Server: https://labelstud.io/guide/storage.html#Local-storage; mount volume in docker container; TODO: better solution for images; volume: `./data:/label-studio/data`
- scp -r steatosis_2024-05-28/ denbi-head:/home/ubuntu/
- ssh denbi-head
- scp -r steatosis_2024-05-28/ node6:/var/git/labelstudio/data/
- Create new project in labelstudio: `Macrosteatosis` and set value in `.env`
- Setup labeling configuration; Settings -> Labeling Interface;
```
```
- Settings -> Cloud Storage -> Add source -> Local files ... -> sync
- Storage title: steatosis_2024-05-28
- Absolute local path: /label-studio/data/steatosis_2024-05-28
- [x] Treat every bucket object as a source file
- Sync storage
- run the `upload_prediction` script
## Annotation task
- open task
- settings:
- [x] show hotkeys on tooltips
- group by label, hide `polygon` to only see points
- annotate keypoints
- select & move to center (dupletten; combined macrosteatosis droplets)
- Select label at bottom (macrosteatosis keypoint 1) and click new
- to save click submit: can be changed later on
- objects for macrosteatosis should be larger then nucleus
## Export labels
Matthias König & Jonas Küttner 2024