PC-YOLO11s: A Lightweight and Effective Feature Extraction Method for Small Target Image Detection.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Small object detection is challenging due to size, resolution, contrast, and background interference.
• Existing YOLO-series models require improvements for effective small object identification.

Purpose of the Study:

• To develop an improved small object detection method, PC-YOLO11s, based on the YOLO11 model.
• To enhance detection accuracy and model efficiency for small objects.

Main Methods:

• Modified YOLO11 network structure by adding a P2 layer for high-resolution feature extraction.
• Removed the P5 layer to reduce computational complexity and model size.
• Integrated a coordinate spatial attention mechanism to improve feature localization.

Main Results:

• PC-YOLO11s achieved superior performance on the VisDrone2019 dataset compared to other YOLO-series models.
• mAP@0.5 increased from 39.5% to 43.8%, and mAP@0.5:0.95 increased from 23.6% to 26.3%.
• Parameter count decreased from 9.416M to 7.103M, indicating improved efficiency.
• Demonstrated strong performance and generalization on tea bud datasets.

Conclusions:

• PC-YOLO11s effectively addresses the challenges of small object detection.
• The model offers significant accuracy improvements and reduced complexity.
• PC-YOLO11s shows excellent generalization ability for practical small object detection applications.