NVIDIA GPU Hackathon - Deep Learning

The ChaLearn LAP team runs a Hackathon on deep learning using Cuda at the Looking at People 2015 ICCV Workshop. During this hackathon, the winning implementations of the two tracks of the Chalearn LAP ICCV Challenge will be provided. The hackathon will consist on a step-by-step training and evaluation of the winning code for this year's winner of the Apparent Age Estimation track at the ICCV competition and a demo of Cultural Event recognition.

Cultural Event Recognition

For this track, we demonstrate the use of CNNs to perform Cultural Event Recognition. The focus is on evaluating the performance differences between CPU and GPU training of Convolutional Neural Networks. You can download the presentation slides here.

Apparent Age Estimation

Rothe, R. Timofte. R. & Van Gool, L. DEX: Deep EXpectation of apparent age from a single image. In ICCV Workshop, 2015

For this track, a full explaination of the training process for the winning code is done. You can find the presentation slides here. For all participants interested on running the demo themselves, the code, data and trained models can be downloaded here:

• Code
• Data
• Trained models
• Readme file

It is also necessary to bring a laptop with the following software installed:

• Matlab
• Python
• Caffe (with Matlab wrapper)

The hackathon will be held on December, 12 at 19:00 hours, at the Grand Hyatt hotel, Santiago de Chile.

We hope to see you there soon,

The ChaLearn LAP team.