In this project, I used some tactics over RGB values of various pixels to detect the black spots on the video. First, video was broken down into frames and then solving the problem for every frame. Goal was to complete the image and removing black square from various frames and then merge them again to form a the new video. Check out some of the frames which have been corrected down below:
PS: 1.I have used plenty of for loops, if you think you can contribute to reduce them! Project is open for contribution and this will improve the code quality! 2. You can apply this tactic to solve simple images like this. This model will brutally fail if used in facial images. Reason: Since I have used average RGB of the surroundings, so it works on simple images like sky or any forest etc.