Unsupervised Learning on Neural Network Outputs
This repo contains the experiment code in paper
Unsupervised Learning on Neural Network Outputs: with Application in Zero-shot Learning
International Joint Conference on Artificial Intelligence (IJCAI) 2016
The paper presents a new zero-shot learning method, which achieved the state-of-the-art results on ImageNet 2011fall (14.2 million images and 21841 classes).
- ILSVRC2012_img_train.tar (138G)
- ILSVRC2012_img_val.tar (6.3G)
- fall11_whole.tar (1.2T)
prepare the images into HDF5 format with
compute the CNN outputs of GoogLeNet of the images with
compute PCA and ICA on the CNN outputs with
compute the MDS features of WordNet graph with
run zero-shot learning experiments with
If you have any question regarding the code and the experiments, please contact me ([email protected]). I would like to hear from you!