search

icatcherplus / icatcher_plus

iCatcher+: Robust and automated annotation of infant gaze from videos collected in laboratory, field, and online studies
GNU General Public License v3.0
17 stars 17 forks source link

Sw/docker capabilities #58

Closed samwinebrake closed 1 year ago

samwinebrake commented 1 year ago

Adds a dockerfile and a script to download models while the image is being created. This PR will allow for automated builds of docker images on docker hub.

yoterel commented 1 year ago

Thanks for this, however I have some issue with the long-term maintenance of this solution: Are container users bound to use icatcher with CPU version only? if so, then maybe my concerns can be dismissed. if not, I am not sure how can this be solved without actively making sure a torch version supporting GPU is installed (I do not have capacity for this). Can you suggest a solution that is future-proof and automatic?

samwinebrake commented 1 year ago

@yoterel , just added a readme. let me know if you think any other instructions would be useful. the Quest is currently working on a repository dedicated to helping with general containerization that will go into more depth about other commands necessary for things such as running on GPU vs. CPU, etc., (might be nice to link from the docker readme in iCatcher+ once complete!) but for now I tried to make it short and sweet!

yoterel commented 1 year ago

Thanks ! Can you fix the other issues (replicated code, and conda env creation) so this can be merged?