Document corner detector like Adobe Scanner with Adobe level accuracy.

We’re looking for an experienced **computer vision / image-processing developer** to build a highly accurate **document boundary and four-corner detection system for uploaded images**. This is **not live camera detection**; the priority is achieving extremely accurate document boundary detection comparable to leading document-scanning apps, including handling uneven backgrounds, shadows, perspective distortion, folded/curved edges, low contrast, and different document sizes. The system should return precise four corner coordinates that can be used for perspective correction and automatic cropping. We’re open to Python, C++, OpenCV, deep-learning models, or a hybrid approach, and the developer should be able to optimize the solution specifically for **accuracy and reliability rather than just basic contour detection**.