CMU's OpenTPOD: Create Deep Learning Object Detectors Without Coding

I'm an OpenTPOD user here at CMU -- used it recently to create a cognitive assistance application. Nice tool -- speeds up the process of creating and tagging videos for object detection training. The team here is looking for suggestions on things to add. PyTorch support is high on my list

I tried the docker-compose install but I'm getting ModuleNotFoundErrors:

    opentpod              |   File "/root/openTPOD/cvat/apps/dataset_manager/task.py", line 24, in <module>
    opentpod              |     from datumaro.components.project import Project, Environment
    opentpod              | ModuleNotFoundError: No module named 'datumaro'