Correctly handle multi-model recordings
Emory provided us with a new input file that duplicated recordings that contained multiple models. The aim of this issue is to ensure that the new input file correctly handles multi-model recordings and that the output at stage 3 (generation of eye-tracker labels) is correct, and preceding steps in the eye-tracking pipeline are correctly handling the new input file and not affecting subsequent steps in the pipeline.
SOLUTION:
CURRENT ACTION PLAN:
- Compare the output of previous stage 3 runs with output post-fix.
- Compare differences between recordings to gain insight into why certain videos are being skipped to help isolate the issue.
- Touch base with Emory to see if we can figure out what is going on with the duplicate recording names.
Edited by afahey2