The implementation of “Higher-order Integration of Hierarchical Convolutional Activations for Fine-grained Visual Categorization” by Sijia Cai, Wangmeng Zuo and Lei Zhang.

• Ubuntu 14.04
• NVIDIA GeForce GTX TITAN X
• MatConvNet (v1.0-beta20)

• First you need to download the FGVC datasets (CUB, Aircraft and Cars) and unzip them into the folder “./datasets”
• Download the pretrained VGG16 model and put it into the folder “./models/pretrained_models”
• Compile MatConvNet with GPU and the cuDNN support
• Train: run hihca_main.m by setting runPhase=train
• Test: run hihca_main.m by setting runPhase=test

If you find the codes of this repository useful, please cite the following paper:

`@inproceedings{cai2017higher,
  title={Higher-Order Integration of Hierarchical Convolutional Activations for Fine-Grained Visual Categorization},
  author={Cai, Sijia and Zuo, Wangmeng and Zhang, Lei},
  booktitle={Proceedings of International Conference on Computer Vision},
  pages={511--520},
  year={2017}
}

`

For any question, please contact

`cssjcai@gmail.com