Wrapper SD-DINO Feature Extractor

A wrapper for feature extractor from A Tale of Two Features that runs over an image dataset, extracts features and applies PCA dimensionality reduction.

Used as a part of DualPM project.

🚀 Installation

Note this requires a separate Python 3.9 environment because of Mask2Former and ODISE dependencies. If using conda: conda create -n feats python=3.9 -y && conda activate feats Setup environment following install.sh

📚 Usage

1. Modify configs/extract_feats.yml to configure load_pca_path, test_root and features_out_root
2. Run: python scripts/extract.py configs=configs/extract_feats.yml

📝 Attribution

Wrapper for: github.com/Junyi42/sd-dino

Please cite the original paper if you use this in your research.

🧑‍💻 Contributors

• Tomas Jakab (tomj@robots.ox.ac.uk)
• Ben Kaye (kaye@robots.ox.ac.uk)