Two-stage deep-learning-based colonoscopy polyp detection incorporating fisheye and reflection correction.

The proposed model can facilitate the clinical diagnosis of colorectal polyps and improve the quality of colonoscopy.