Zero-shot learning methods for ImageNet:
- CADA-VAE (remplementation, see the original implementation):
- SimpleZSL (based on DeViSE)
Config files in "config" folders are just examples. Config files and models used in our experiments are available to download: Config files and models
aux_feats
files with encoded class auxiliary (text) features were from: aux_feats
pickle files (See Encoded Wikipedia articles (extracted features))
conda.yml contains a Conda environment used for the project. Note that it contains more dependencies than this project requires!
The project requires to have subprojects directories in PYTHONPATH
:
export PYTHONPATH=$PYTHONPATH:${PROJECT_DIR}:${PROJECT_DIR}/CADA-VAE:${PROJECT_DIR}/SimpleZSL
The code from this repository was used in our work, see our project page.