# Acupuncture Point Location Detection System - Abstract

The Acupuncture Point Location Detection System is an innovative deep learning-based computer vision project that addresses the precision challenges in traditional Chinese medicine acupuncture therapy by integrating infrared imaging technology with advanced deep learning algorithms. Built on a ResNet18 backbone with innovative modules including residual connections, domain adaptation layers, and attention mechanisms, the system achieves over 85% detection accuracy with positioning error less than 5 pixels while operating at 30 FPS with less than 100ms latency for clinical applications. The core technology features multi-modal fusion of infrared and visible light imaging, APAP-based image registration for precise alignment, and advanced image processing techniques enabling sub-pixel detection with adaptive thresholding. Utilizing a specialized dataset of Spleen Yu acupoint (脾俞穴) with high-resolution images (1920×1080) from multiple subjects, the system employs comprehensive data augmentation strategies and achieves convergence through optimized loss functions and learning rate scheduling. Built with PyTorch framework and featuring GPU acceleration support, the project has completed comprehensive algorithm documentation and is ready for implementation and testing, representing a significant step toward modernizing traditional Chinese medicine through artificial intelligence technology with applications in clinical practice, medical education, and research development.