Just realized detection part. Updating...

Requirements

• python2.7
• pytorch>=0.4
• numpy

Quickstart

You can use the optional argument -h to see more arguments you can set.

• Generate the dataset: $ python readData.py
• Train the model: $ python train.py . Set some parameters as you want.
• Test: $ python activities_prediction_srnn.py . The outputs are the detection/prediction accuracy.

Neural network

Based on the original codes and the paper, the structure of this neural network is following: