lijuny/low-rank-bilinear-pooling

Fine-grained classification via second order statistics in a compact end-to-end trainable model

Low-Rank-Bilinear-Pooling

Code, demo and model for our project of low rank bilinear pooling for fine-grained classification

For papers, slides and posters, please refer to our project page

Several demos are/will be included as below --

1. demo 1: quick training using caffe, including matlab files for initialization [available];
2. demo 2: hyperparameter study by low-rank and co-decomposition on the classifier parameters [available];
3. demo 3: three methods of visualization [available];
4. demo 4: fine-tuning the network using matconvnet [available];

For each demo, there will be some models involved. Please download those models from the google drive

The folder named "caffe-20160312" is a modified caffe toolbox by Yang Gao for his "compact bilinear pooling" page. Please follow the caffe instructions for compiling the toolbox. The configuration file within it should be revised accordingly.

As for details on the training, demo and code, please go into each demo folder.

If you find our model/method/dataset useful, please cite our work:

@inproceedings{kong2017lowrankbilinear,
  title={Low-rank Bilinear Pooling for Fine-Grained Classification},
  author={Kong, Shu and Fowlkes, Charless},
  booktitle={CVPR},
  year={2017}
}

last update: 05/09/2017

Shu Kong

aimerykong At g-m-a-i-l dot com