This allows to preprocess (extract patches or entire face), and apply the quality check to the preprocessed data. If quality is bad, the preprocessed sample is not saved.
More specifically you can run a code:
./bin/spoof.py \
celeb-a \
lbp-svm \
--skip-extractor-training --skip-extraction --skip-projector-training --skip-projection --skip-score-computation --allow-missing-files \
--grid idiap \
--groups train \
--preprocessor rgb-face-detect-check-quality-128x128 \
--sub-directory <path_to_save>It will detect the face in the CelebA, and check the quality, by applying eye-detection and estimating if detected eyes are located in the expected positions.