EB-YOLO:An efficient and lightweight blood cell detector based on the YOLO algorithm.

Area of Science:

• Medical diagnosis
• Computer vision
• Biomedical engineering

Background:

• Blood cell detection is crucial for medical diagnosis.
• Current high-precision object detection models are computationally expensive for low-end devices.
• Lightweight models offer speed but lack accuracy in complex blood cell detection tasks.

Purpose of the Study:

• To develop an efficient and highly accurate blood cell detector for resource-constrained environments.
• To address the limitations of existing models in terms of speed and accuracy for real-time blood cell analysis.

Main Methods:

• Proposed an Efficient Blood Cell Detector based on YOLO (EB-YOLO).
• Utilized ShuffleNet as the backbone for efficient feature extraction.
• Incorporated Convolutional Block Attention Module (CBAM) and Adaptive Spatial Feature Fusion (ASFF) for enhanced feature representation and integration.
• Employed depth-wise separable convolution to reduce model parameters and computational load.

Main Results:

• Achieved 92.1% mAP@50 on the BCCD dataset.
• Demonstrated low computational complexity (0.9 GFLOPs) and a small parameter count (0.289M).
• Showed superior inference speed compared to classic YOLO on Raspberry Pi 5.

Conclusions:

• The EB-YOLO model effectively balances lightweight design with high accuracy for blood cell detection.
• The proposed method shows significant promise for deployment on low-end embedded systems for real-time medical diagnosis.