Wardrobe Vision Engine is a real-time fashion and garment computer vision workspace: YOLO detection, Lab colour analysis, SAM / FastSAM / SAM3, QR and barcode reading, a tkinter GUI, and training pipelines for DeepFashion2 and ModaNet.
Repository: github.com/hmazin/Wardrobe_Vision_Engine
git clone https://github.com/hmazin/Wardrobe_Vision_Engine.git
cd Wardrobe_Vision_Engine # or your local folder name
# Python 3.10+ recommended
python -m venv .venv && source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e .
# Optional extras (GPU PyTorch from pytorch.org as needed)
pip install -e ".[all]"GPU: install a CUDA build of PyTorch first, then pip install -e .
Conda: see docs/setup/conda.md and environment.yml (env name: wardrobe-vision-engine).
# Graphical launcher (recommended)
wardrobe-vision-gui
# CLI helpers
wardrobe-vision-detect basic
wardrobe-vision-detect garment
wardrobe-vision-train deepfashion2
# Or run a module directly
python -m wardrobe_vision_engine.detect.basicMore detail: docs/quickstart.md · docs/index.md.
Wardrobe_Vision_Engine/
src/wardrobe_vision_engine/ # Python package (only copy of the code)
configs/datasets/ # Ultralytics dataset YAML files
docs/ # User and setup documentation
tests/ # pytest
scripts/ # e.g. upload_weights_to_hf.py
Entry points are defined in pyproject.toml ([project.scripts]). There are no duplicate detect_*.py files at the repository root.
| Variable | Default | Description |
|---|---|---|
WVE_DATASETS_ROOT |
~/datasets |
All datasets (DeepFashion2, ModaNet, …) |
WVE_RUNS_ROOT |
~/runs |
Ultralytics training output |
YOLO_MODEL_PATH |
(varies) | Weights for detection scripts (GUI sets this) |
RECORD_VIDEO |
0 |
Set to 1 to record the session |
RECORD_VIDEO_PATH |
output.mp4 |
Output video path |
Legacy: TGI_DATASETS_ROOT / TGI_RUNS_ROOT are still read if WVE_* is unset.
pip install -e ".[dev]"
pytest
ruff check src tests
black --check src testsSee CONTRIBUTING.md.
MIT. Third-party datasets and model weights (DeepFashion2, ModaNet, COCO, SAM, YOLO) remain under their own licenses; do not redistribute dataset files in this repository.