Collaborative Computer Vision and Cloud Platform for Gastrointestinal Polyp Detection

Gastric cancer is a common malignant tumor in the world. Regular endoscopic examination and timely removal of precancerous polyps in early patients can significantly reduce mortality. A real-time detection model for gastrointestinal polyps based on the improved YOLOv7 algorithm is proposed to address the issues of unclear early symptoms, small target size, high missed and misdiagnosis rates of gastric cancer. Firstly, a comprehensive and intuitive endoscopic polyp image dataset was constructed by collecting publicly available gastrointestinal polyp image data; Then, based on the YOLOv7 algorithm, the CBAM attention mechanism is introduced to improve the algorithm’s feature extraction ability in complex situations, and the WIoU bounding box regression loss function is used to improve the positioning accuracy of the target bounding box. The experimental results show that the improved model accuracy has increased from 88.8% to 92.5%, and the average accuracy has increased from 91.25% to 92.16%, which can better meet the real-time accuracy and speed requirements of endoscopic examination. Finally, deploying the trained model to the cloud service platform can provide free testing services to the public conveniently and flexibly.