MRCNN

Implementation of Manifold Regularized Convolutional Neural Networks.

Dependency

• Numpy
• Scipy
• Tensorflow >= 1.0
• Pretrained VGG Net
• CIFAR-100 Dataset

Setup

Run bash setup.sh in terminal to automatically download the CIFAR-100 dataset and pretrained VGG19 model.

Usage

Enter python MRCNN.py in bash for fast with default setting.

Use --gpu_id to specify the GPU devices to use. By default, all GPUs available will be used.

Use --vgg_path to specify the path of pretrained VGG Net. By default, the model is located under vgg/ directory

Use --result_path to specify file to save the result. By default, the file is "result.csv".

Use --help to acquire more information.

Citation

If you'd like to compare our framework with your work, please cite our paper as following:

TBA