Recurrent Convolutions

Excited to be a finalist for the OpenPower developer challenge!

Posted today at 6:42 pm

OpenPOWER Foundation is an ambitious effort led by IBM, Nvidia and others to build open data center technologies.

OpenPOWER had run a developer challenge on http://openpower.devpost.com with Deep Learning as one of their tracks. Participating in this challenge was a very great experience for me. Especially the part where they provided all the participants with a 32GB RAM OpenPOWER instance with an Nvidia Tesla K80 GPU with 12 gigs of VRAM to crunch data in a breeze!

For the challenge, I developed a Deep Learning system which can identify, classify and organize Paintings on the basis of their Genre. I sourced these Paintings from a Kaggle contest data who in turn sourced it from http://www.wikiart.org

I’ll be posting a tutorial on how I approached this project in a couple of weeks and release the code as well. (Right now the code is a total mess! :P ) I’m planning to extend the system to identify the Painter as the Kaggle challenge requires.

To my surprise, our submission got into the finals! My younger brother, though just a beginner, gave me great inputs! Here’s the announcement they made :


Written by Praveen Sridhar who lives in Kochi, India and works on Machine Learning projects in Python using the wonderful scikit-learn, TensorFlow and Keras libraries. Here's where to find him on Twitter & Github. I've been invited for Deep Learning School! Seeking your support at http://contribute.recurrentconvolutions.com