Image Inpainting Using Deep Learning Architectures: PConv and GAN

Image Inpainting is an image reconstruction technique, where missing sections of an image are filled or restored. Image inpainting is now one of the emerging technique in the field of image processing. It also has various applications in computer vision, where images are used for recognition or detection. The focus of this work is to address the challenge of reconstructing damaged or missing regions in images, a task crucial for various applications ranging from image editing to forensic analysis. In the proposed method, Partial Convolution (PConv) and Generative Adversarial Network (GAN) method are used to train the model on the features extracted using encoder and a random mask is generated over image for inpainting. The proposed model is tested on CelebA Dataset. The model has achieved 0.82 SSIM on PConv model and 0.84 SSIM on GAN. However GANs produced better results than PConv. The future of image inpainting holds exciting possibilities and potential advancements that could significantly impact various domains.