Project to implement image inpainting on masked images.
Used https://www.kaggle.com/datasets/theblackmamba31/landscape-image-colorization, kaggles landscape images for training. Created dataset for this specific project, by applying masks, randomly generated by opencv on downloaded images.
Frameworks/libraries used: -Pytorch to make the complete GAN model (generator and discrimminator), define reconstruction loss function, train using pixelwise and adversarial loss. -OpenCV To deal with all the images, apply masks to images an downloaded dataset -Matplotlib and Seaborn To plot the graphs of generator and discriminator loss functions wrt number of iterations passes -Numpy To perform mathematical operations on the image array like, creating masks, multiplying masks with predictions etc
Detailed explanation for the same can be found in Project_abstract file