The Deep learning model developed fills the distorted parts of the image caused due to some occlusions like window panes, fencing or dirt on the camera. We filled those parts using Convolutions and Transposed convolutions utilizing Tensorflow, OpenCV, Numpy in python which removes the unwanted objects and fills out the original image which would have been there without the occlusions.
The Dataset required for training was downloaded from multiple sources and was pre-processed by us using libraries like OpenCV and Numpy.
Here are few Example samples from the test dataset
More details about the implementation can be found in the report.