Code and trained Convolutional Neural Networks for emotion recognition

• Python notebook with example usage.
• Zip file with code for converting RGB values to our mapped LBP codes.
• Link to public dropbox zip file with trained CNN models. The models include VGG_S trained on RGB and the four mapped LBP-based representations described in the paper. Warning, ~2GB file!
• A Gist page for our trained models, now appears in the BVLC/Caffe Model Zoo.
• For more information see the project page.

Code and trained Convolutional neural networks for age and gender classification

• python notebook with example usage.
• A git repository with various scripts and definition files to help reproduce our results.
• Caffe model files and additional data in the zip file cnn_age_gender_models_and_data.0.0.2.zip (~86MB)
• A Gist page for our trained models, now appears in the BVLC/Caffe Model Zoo.
• For more information see the project page or blog post.

The LATCH binary descriptor

Our LATCH descriptor has been fully merged into OpenCV 3.0:

• latch.cpp at OpenCV
• The full, merged pull request is available from this link.
• Example usage, also merged into OpenCV is available in LATCH_match.cpp
• OpenCV xfeatures2d page including LATCH and many other 2D image descriptors.
• CUDA (GPU) implementation of the LATCH descriptor and a cool visual odometry demo both by Christopher Parker.
• For more information see the project page or blog post.