Reports on Clinical Oncology from University of Coimbra Provide New Insights (Highly Accurate and Fast Yolov4-based Polyp Detection).

A research study conducted at the University of Coimbra in Portugal has explored the use of deep learning models to improve the detection of polyps, which are precursors to colorectal cancer (CRC). The study utilized the YOLOv4 model and deployed it on NVIDIA TensorRT to optimize and quantize the model for real-time performance. Various techniques such as data augmentation and transfer learning were tested to enhance the accuracy of the model. The research found that quantization, specifically at the FP16 and INT8 levels, did not significantly degrade the model's accuracy and could potentially be used as a regularization technique. The study also highlighted the importance of data augmentation in overcoming the lack of labeled data in the medical field. [Extracted from the article]