sae-mxnet
A python implementation for "Semantic Autoencoder for Zero-Shot Learning"
Inspired by https://github.com/hoseong-kim/sae-pytorch
Use raw images and semantic features from AwA2 dataset instead of .mat file.
Use Res101 pretrained model from mxnet for image feature extraction.
When HITK=1, the accuracy will be 80.97%