Professional segmentation annotation for DeepFaceLab with SAM integration, polygon tools, and real-time mask editing
Precision painting with adjustable brush sizes and polygon drawing for complex shapes
Segment Anything Model auto-masking for rapid annotation with AI assistance
30-step history with Ctrl+Z/Y support for seamless workflow and error correction
Smooth zooming with mouse wheel and right-drag panning for detailed work
Erode, dilate, and feather operations with adjustable kernel sizes
Save masks as PNG and export polygon annotations as LabelMe JSON format
# Run the scaffold script
./scripts/scaffold_annotator.sh
# Install dependencies
pip install -r requirements-annotator.txt
# Optional: Install SAM dependencies
pip install torch torchvision
pip install git+https://github.com/facebookresearch/segment-anything.git
# Start the annotator
./scripts/run_annotator.sh
# With SAM enabled
./scripts/run_annotator.sh --sam --sam-checkpoint /path/to/sam_vit_h.pth
# With auto-fill on load
./scripts/run_annotator.sh --sam --sam-checkpoint /path/to/sam_vit_h.pth --sam-autofill
Place your source images in the extract_frames directory before starting. The annotator will automatically detect them.
# Custom directories
./scripts/run_annotator.sh \
--frame-dir /path/to/your/frames \
--mask-dir /path/to/your/masks \
--ann-dir /path/to/your/annotations
Start annotating your face segmentation masks with professional tools and AI assistance