The problem Capture It Openly solves
Current subject removal techniques are very tedious because a user has to select a subject by drawing a mask manually, whereas our model creates the mask automatically and the user just have to select the subject he/she want to remove which is very efficient.
Challenges we ran into
- We were not familiar with UNET architecture and CGANS ,so learning it was quite a tedious task
- Tuning the hyper parameters to reduce loss during the learning process
- Finding the correct dataset for the semantic segmenation and CGANS was the hardest part for us