Low-cost real-time skeleton key point recognition method and device

Abstract

The invention discloses a low-cost real-time skeleton key point recognition device, which comprises an image acquisition module, a core calculation unit, a lightweight neural network algorithm module, a neural network acceleration engine and a skeleton key point output module, wherein the image acquisition module acquires images; sends the acquired image information to a core computing unit; the core calculation unit performs image processing on the acquired image, the lightweight neural network algorithm module adopts an improved SqueezeNet as a basic backbone network, performs multi-scale feature extraction on the image in combination with a feature pyramid network, accelerates the network through the neural network acceleration module, and performs multi-scale feature extraction on the image. Finally, skeleton key points are output through a skeleton key point output module, any monocular camera is adopted by an image acquisition module, and a mobile terminal CPU is adopted by a core calculation unit.

Description

Technical field [0001] The present invention relates to the field of computer visual techniques, and more particularly to a low-cost real-time bone key identification method and apparatus. Background technique [0002] Skeleton key identification technology is one of the basic technology of computer vision. The technique detects the articular joints, facial features in image / video data, by sensors (camera, infrared), and a five senses, and the human bone information is described by key points. [0003] However, existing deep learning-based bone-critical identification algorithms are difficult to run in real time in a low-cost hardware platform, need to match high cost hardware (such as GPU or high-end cameras) to achieve real-time. The current invention is based on a series of hardware and software optimization techniques, and real-time identification of bone key points can be completed on a low-cost hardware platform. Under the premise of ensuring accuracy, about 50 millisecon...

AI Technical Summary

Problems solved by technology

[0004] The traditional bone key point algorithm is based on the idea of ​​template matching on the basis of geometric prior, and the accuracy is poor.

Due to the limitation of hardware performance, the existing bone key point recognition algorithm based on deep learning has a slow recognition speed on low-cost hardware platforms (such as mobile phones and tablets), and the algorithm linkage application will cause the application to freeze and lose frames. and so on, which greatly affects the user experience

Images