MvNNcor

This is an implementation of Deep Embedded Complementary and Interactive Information for Multi-view Classification (MvNNcor) in Pytorch.

Requirements

• Python=3.5.6
• Pytorch=1.0.0
• Torchvision=0.2.1

Datasets

The model is trained on AWA/Caltech101/Caltech20/NUSOBJ/Reuters/Hand dataset, where each dataset are splited into three parts: 70% samples for training, two-thirds of the rest samples for validation, and one-third of that for testing. We utilize the classification accuracy to evaluate the performance of all the methods.

Implementation

# Train the model on AWA dataset
python MvNNcor_Train.py --dataset_dir=./mvdata/AWA/Features --data_name=AWA --num_classes=50 --num_view=6 --gamma=6.0

# Test MvNNcor on AWA dataset
python MvNNcor_Test.py --dataset_dir=./mvdata/AWA/Features --data_name=AWA --resume=./results/.../model_best.pth.tar --num_classes=50 --num_view=6 --gamma=6.0

Citation

Deep Embedded Complementary and Interactive Information for Multi-view Classification. AAAI2020