- src, model training, post processing and evaluation code
- data, gallery and query features using SAKE model with distribution alignment loss
-
(In the repo root dir) Calculate and save scores.
python src/pre_process.py
-
Run cluster and retrieval experiments。
-
KMeans w/ fuse
python src/clusterretr.py --feature_file ./data/features_processed.pickle > results/kld_sketchy_fuse0.2.log --fuse 0.2
-
KMeans w/o fuse
python src/clusterretr.py --feature_file ./data/features_processed.pickle > results/kld_sketchy.log