Open SabolSocare opened 2 weeks ago
makseq
commented
2 weeks ago How does your docker-compose.yml look?
Did you follow step #1 from https://github.com/HumanSignal/label-studio-ml-backend/blob/master/label_studio_ml/examples/yolo/README.md#quick-start ?
SabolSocare
commented
2 weeks ago Here's my Docker Compose configuration for the YOLO service:
It works when the image is uploaded from the browser, but it does not work with images synced from cloud storage.
version: "3.8"
services:
yolo:
container_name: yolo
image: humansignal/yolo:v0
build:
context: .
args:
TEST_ENV: ${TEST_ENV}
environment:
# Optional basic auth for the model server
- BASIC_AUTH_USER=
- BASIC_AUTH_PASS=
# Logging and server configuration
- LOG_LEVEL=DEBUG
- WORKERS=1
- THREADS=8
- MODEL_DIR=/data/models
- PYTHONPATH=/app
# Label Studio integration
- LABEL_STUDIO_URL=http://127.0.0.1:8089
- LABEL_STUDIO_API_KEY=9e06b3eca8f7edfc0f9b9XXXXXXXXXX
# YOLO-specific configuration
- ALLOW_CUSTOM_MODEL_PATH=true
- DEBUG_PLOT=false
- MODEL_SCORE_THRESHOLD=0.5
- MODEL_ROOT=/app/models
ports:
- "9090:9090"
volumes:
- "./data/server:/data"
- "./models:/app/models"
- "./cache_dir:/app/cache_dir"
LABEL_STUDIO_URL=http://127.0.0.1:8089
This line is the same as the use of localhost. When you are inside of the docker, it won't work. In the example docker compose configuration it was a comment about label studio url, please read it carefully. https://github.com/HumanSignal/label-studio-ml-backend/blob/master/label_studio_ml/examples/yolo/docker-compose.yml#L26
SabolSocare
commented
1 week ago Thank you for the guidance. The URL http://127.0.0.1:8089 was just an example; I’m actually using my own domain, which works (auto-annotation) as expected with data under /data/upload. However, I'm encountering issues specifically with data paths prefixed by /data/local-files/? (Cloud Storage). It seems that these paths aren't accessible in the same way
I am unable to use the Yolo ML Backend model to annotate image data from Cloud Storage.