Computer Vision Document Scanner

I developed a document scanning pipeline that automatically detects the largest quadrilateral contour in an image, applies perspective transformation to correct skew and angle, and outputs a flat, scan-like version of the document. The algorithm begins with grayscale conversion, Gaussian blur, and Canny edge detection, then finds contours and selects the four-point polygon with maximum area. Using OpenCV's warp perspective, the selected region is mapped to a top-down view with adjustable output resolution. I wrapped the entire pipeline into a simple Gradio web interface, allowing users to upload photos from their phone and receive corrected scans instantly. This project deepened my understanding of image preprocessing, geometric transformations, and building accessible computer vision tools for non-technical users.