I require the development of a web-based background removal tool to remove backgrounds from images automatically. The tool should utilize state-of-the-art deep learning techniques like semantic segmentation and instance segmentation to identify and isolate foreground objects from their backgrounds with high accuracy. It is important that the extracted foregrounds are as clean as possible with no visible artifacts or remnants from the original background. The web app should have an easy-to-use interface where users can upload an image, select the foreground objects they want to retain, and remove the background with a single click. Advanced customization options like adjusting eraser size, threshold, etc. would be appreciated but not necessary. Once processed, the background-removed images should be available for instant download in PNG format. Integration with cloud storage for handling large files and version control would be beneficial. Experience with TensorFlow, PyTorch or other deep learning frameworks for computer vision tasks and building full-stack web applications is required to take up this challenging yet interesting project.
