terrypang / bupt-dr2

0. Environment

• Intel I7-7655K, 32G DDR4, Nvidia Titan X
• Xubuntu 16.04.1, CUDA 8.0, cudnn 5.1

1. Install the following components

• python 2.7
• numpy
• scipy
• pandas
• matplotlib
• scikit-learn
• pillow
• tensorflow
• keras
• h5py
• pydot

2. Folder structure

├── data
│   ├── raw
│   ├── test
│       └── test
├── features
├── logs
├── models
├── submitions
├── utils
├── weights
├── submit.py
└── train.py

2. Download model weight files to 'weights/imagenet'

• https://github.com/fchollet/deep-learning-models/releases/download/v0.4/xception_weights_tf_dim_ordering_tf_kernels_notop.h5
• https://github.com/fchollet/deep-learning-models/releases/download/v0.2/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5
• https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5

3. Download kaggle compitition data

• Download from: https://www.kaggle.com/c/dogs-vs-cats-redux-kernels-edition/data
• Extract all images from train.zip to 'data/raw'
• Extract all images from test.zip to 'data/test/test'
• Copy 'sample_submission.csv' to 'submitions'

4. Train single models and get results

• Run 'python train.py -m [inception|resnet50|xception]' to get transform learning model
• Run 'python finetune.py -m [inception|resnet50|xception]' to fine tune model
• All model weight files are in the folder 'weights'
• Run 'python submit.py -m [inception|resnet50|xception]' to get submition
• All submitions are in the folder 'submitions'

5. Train blend model and get results

• Run 'python blend.py -m xception resnet50 inception' to get blend result with validation
• Run 'python blend.py -m xception resnet50 inception -a true' to get blend result without validation