Using Convolutional Neural Networks to Analyze and Detect Key Points of Objects in Image

The article discusses the use of Convolutional Neural Networks (CNNs) for analysis and detection of key points of objects in images, in particular hands. The mathematical representation of the principles of CNNs work explains the keypoint detection process, and the conducted experiments show the significant influence of lighting conditions and frame rate (frame per second - FPS) on detection accuracy and processing delay. The results confirm that increasing illumination up to 500 lux and frame rates up to 60 FPS improve accuracy and reduce latency, although detection accuracy decreases in low light. The findings highlight the importance of lighting and frame rate optimization to achieve high real-time performance.