I have tried to solve the problems for the last 2 days however i have faced myriad of problems regarding the baselines:
- The datasets are too large and take significant time to download which eats into collab time and disk space. (Not a huge problem but a huge inconvenience)
- OBJDE baseline doesnt work if all images dont have a single output as a prediction and straight up get a 0
- IMGCOL baseline always reports an image missing for which I have no clue despite ensuring that the image is present in the submission.
All these bugs make it very hard to remain motivated and waste a lot of time. Please look into it ASAP as the deadline is very close