This is a translation into Caffe of Anil Thomas's Neon-based project for Kaggles Whale Detection Challenge.

• Competition page at Kaggle
• Anil's Code
• Caffe

This is a work in progress and most of the code will still require some of the Python scripts from Anil's repo

1. Download and install Caffe. Check instructions here

2. Create /train, /test and /traincrops directories using Anil's repo. All you need is call prep method in run.sh in his repo. Copy all three directories to data/ directory here. Also download sample_submission.csv file to data/.

3. Declare CAFFE_ROOT environmental variable.

   export CAFFE_ROOT=/path/to/caffe/folder
   

4. Call run.sh

   ./run.sh
   

1. Need to write wrapper scripts to predict Bonnet and Head for test images; and crop test images (possibly using anlthm/whalw-2015/crop.py).

Thanks Anil Thomas and Team Nervana for initial code. This project has started as a re-implentation of original code. Special Thanks to Praveen (https://github.com/opraveen) For developing critical components of the code.