Skip to content
This repository has the pytorch implementation of the paper 'Generalized Zero- and Few-Shot Learning via Aligned Variational Autoencoders' (CVPR 2019)
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


This repository has the pytorch implementation of the paper "Generalized Zero-and Few-Shot Learning via Aligned Variational Autoencoders." (CVPR 2019) [pdf].

This repository has the implementation of zero-shot learning in a Generalized setting and has been tested on 4 datasets.

Note: I am still working on improving the results.


The dataset splits can be downloaded here, please download the Proposed Split and place it in the same folder.

Find additional details about the dataset in the of the Proposed split.

Training and Testing

Download the pretrained model for various datasets [here] and place it in models/

  1. For Testing:
python --dataset CUB --dataset_path xlsa17/data/CUB/ --model_path models/checkpoint_cada_CUB.pth --pretrained

Change the arguments according to the dataset

  1. For Training:
python --dataset CUB --dataset_path xlsa17/data/CUB/


Dataset Paper Results
(s, u, h)
Respository Results
(s, u, h)
CUB 53.5, 51.6, 52.4 53.52, 47.29, 50.21
AWA1 72.8, 57.3, 64.1 73.54, 46.69, 57.19
AWA2 75.0, 55.8, 63.9 82.77, 44.94, 58.25
SUN 35.7, 47.2, 40.6 39.03, 37.43, 38.21
You can’t perform that action at this time.