Deep Learning for Real-Time Malaria Parasite Detection and Counting Using YOLO-mp

Malaria remains a life‑threatening disease, and accurate quantification of Plasmodium parasites in blood smear microscopy is essential for diagnosis and treatment monitoring. This work presents a deep learning framework for automated detection and counting of malaria parasites from microscopic images. Two state‑of‑the‑art object detection architectures are independently trained and evaluated on a public YOLO‑formatted dataset containing 1,693 training, 360 validation, and 365 test images. The first model enhances YOLO26 with a Convolutional Block Attention Module (CBAM) to refine feature representations. The second employs RT‑DETR, a transformer‑based detector that eliminates non‑maximum suppression via bipartite matching. Both models undergo rigorous evaluation using COCO metrics (mAP@0.5, mAP@0.5:0.95, precision, recall). A complete web application built with Flask, HTML, CSS, and JavaScript integrates the trained models, allowing users to upload blood smear images, receive parasite counts with annotated bounding boxes, and obtain clinical burden classification (negative, low, moderate, high, very high). The system achieves robust detection performance while providing an accessible interface for malaria parasite counting, supporting clinicians in rapid and consistent assessment.

Keywords: Malaria detection, Plasmodium, YOLO26, CBAM, RT‑DETR, object detection, parasite counting, deep learning, medical image analysis, Flask web application.