experiments

experiments

This folder holds python code used for the experimental section.

• actions holds code for predicting actions given the predicted map (supplementary material)
• assessment holds code to run the visual assessment experiment (sec 5.4 of the paper)
• dataset_stats holds code to measure some data-related statistics.
• metrics holds code to compute metrics from precomputed predictions.
• mlnet_comparison holds code to train and predict a sequence from Multi Level Network from Cornia et al.
• predict_on_cineca holds code to predict sequences with the dreyevenet model.
• rmdn_comparison holds code to extract C3D features, train and predict using the Recurrend Mixture Density Network from Bazzani et al.
• train holds code to train the dreyevenet model.
• visualization holds code for visualizing some predictions (mainly used for paper figures)

All python code has been developed and tested with Keras 1 and using Theano as backend.