Yolov8 Detection Label Studio Active Learning
Setup
cd LabelStudio-Yolov8-Detection-Backend
pip install -r requirements.txt
- Set environment variables
export LABEL_STUDIO_ML_BACKEND_V2=True
- Set path in yolov8.py, use absolute path
# labelstudio annotation: "image": "/data/upload/20/image1.png" or "image": "/data/local-files/?d=LABEL_STUDIO_LOCAL_FILES_DOCUMENT_ROOT/image1.png"
LABELSTUDIO_PATH = '/data/upload/20'
DATASET_ROOT_DIR = '/home/user/apps/label-studio/media/upload/20' # folder where your images is store
STORAGE_DIR = '/home/user/LabelStudio-Yolov8-Detection-Backend/data' # folder to save training artifacts
MODEL_NAME = 'yolov8n_custom' # name your model
MODEL_INIT = '/home/user/LabelStudio-Yolov8-Detection-Backend/model/yolov8n.pt' # model init
MODEL_LATEST = f'/home/user/LabelStudio-Yolov8-Detection-Backend/runs/detect/{MODEL_NAME}/weights/best.pt' # latest model updated after each training
DETECTION_YAML = '/home/user/LabelStudio-Yolov8-Detection-Backend/data/detection.yaml' # your yaml for training yolov8
- Setup yolov8 dataset yaml in data/detection.yaml
train: /home/user/LabelStudio-Yolov8-Detection-Backend/data/images
val: /home/user/LabelStudio-Yolov8-Detection-Backend/data/images
nc: 2
names: ['table', 'signature']
Create - launch - connect ML backend server
label-studio-ml init yolov8_backend --script yolov8.py --force
label-studio-ml start yolov8_backend
- Connect Label Studio with ML backend
