This repo contains the source code for the following papers:
Structural-RNN (S-RNN) for doing deep learning over spatio-temporal graphs. Checkout the srnn branch and the source code is present in the directory structural_rnn. This source code takes input a spatio-temporal graph written in a text file and creates the S-RNN architecture. Currently the directory contains the code for human motion modeling on H3.6m data set. Our paper can be downloaded from here
Sensory-fusion recurrent neural architecture for driver activity anticipation http://brain4cars.com The source code is in the directory anticipatory-rnn/maneuver-rnn. Checkout 'icraversion' branch of NeuralModels to reproduce our results.
S-RNN for modeling human-object interaction (human activity detection and anticipation) code is present in anticipatory-rnn/activity-anticipation. Our paper can be downloaded from here