Interactive Object Removal and Inpainting System Using Deep Learning

This paper presents an interactive system for object removal and image inpainting that combines Mask R-CNN for object segmentation and DeepFillv2 for image completion. The system en-ables users to select unwanted objects through a bounding box interface, automatically segments the selected objects using a ResNet50-backed Mask R-CNN architecture, and seamlessly recon-structs the background using DeepFillv2’s generative adversarial network. Our approach addresses the challenges of manual image editing by providing an intuitive user interface while leveraging state-of-the-art deep learning models for accurate object removal and realistic image completion. Experimental results demonstrate the system’s effectiveness in maintaining visual coherence while removing user-selected objects from various image types.