This study explores the application of deep learning methods for the detection of colorectal polyps, a precancerous lesion. The authors aim to develop a highly accurate detection model by analyzing and comparing leading target detection algorithms, including FCOS, YOLO, and Faster RCNN. The research leverages deep neural networks to enhance diagnostic accuracy and efficiency in early detection of colorectal polyps, which is crucial for preventing colorectal cancer. The study uses a comprehensive dataset of CT scans from patients with colorectal polyps and evaluates various RCNN-based detection algorithms to identify the most effective model parameters. The experimental design includes feature extraction using ResNet50, ResNet101, and ResNeSt50, with a focus on model convergence and accuracy. The results indicate that two-stage object detection models outperform single-stage models, and deeper feature extraction networks, such as ResNeSt101, yield better performance. The study concludes that the proposed model significantly improves the detection accuracy of colorectal polyps, contributing to more precise and efficient early diagnosis and prevention of colorectal cancer.This study explores the application of deep learning methods for the detection of colorectal polyps, a precancerous lesion. The authors aim to develop a highly accurate detection model by analyzing and comparing leading target detection algorithms, including FCOS, YOLO, and Faster RCNN. The research leverages deep neural networks to enhance diagnostic accuracy and efficiency in early detection of colorectal polyps, which is crucial for preventing colorectal cancer. The study uses a comprehensive dataset of CT scans from patients with colorectal polyps and evaluates various RCNN-based detection algorithms to identify the most effective model parameters. The experimental design includes feature extraction using ResNet50, ResNet101, and ResNeSt50, with a focus on model convergence and accuracy. The results indicate that two-stage object detection models outperform single-stage models, and deeper feature extraction networks, such as ResNeSt101, yield better performance. The study concludes that the proposed model significantly improves the detection accuracy of colorectal polyps, contributing to more precise and efficient early diagnosis and prevention of colorectal cancer.