Classify behaviours and superimpose the results on video
This program takes an input video, and tracking data derived from CppMT (https://github.com/gnebehay/CppMT) or OpenFace (https://github.com/TadasBaltrusaitis/OpenFace). The user classifies randomly chosen frames from the period of interest by selecting a numeric value corresponding to each behaviour that is being encoded (e.g. 1 for looking at an iPad, 0 for looking at a TV). Once an appropriate number of frames have been classified, the code attempts to predict the behaviours being observed in the remaining frames in the video.
Given an input video, outputvideo and 1+ bounding-box/prediction files, this program will superimpose (a) box(es) with the coordinates given in the bounding-box/prediction file(s), whose colour depends on the predicted state for each frame included in the file. It is used to check the predictions from, e.g. abc-classify.py
The IDInteraction Processing Pipelines were developed in the IDInteraction project, funded by the Engineering and Physical Sciences Research Council, UK through grant agreement number EP/M017133/1.
Copyright (c) 2017 The University of Manchester, UK.
Licenced under LGPL version 2.1. See LICENCE for details.