Deployment and Implementation of Facial Feature Recognition Algorithm Based on RP-rv1126 Development Board
Chao Li
Cethik Group Co.,Ltd
Linlu Zou
City college, Zhejiang University
DOI: https://doi.org/10.59429/esta.v11i3.7332
Keywords: Facial Feature Recognition; RP-rv1126; RKNN; NPU; Algorithm deployment
Abstract
Firstly, it is necessary to use the Rockchip Micro RKNN tool to convert the inference model into an RKNN model that can run on edge devices. At the same time, a C++main program should be written to facilitate calling the RKNN model for feature inference on specific images; Configure the compilation options for the development board, enable the NPU mini driver, and after successful image compilation, generate an image file with the NPU driver enabled and re burn the development board. Finally, by uploading the main program, rknn inference model, and the image to be inferred to the development board, the inference of the algorithm on the development board can be achieved. Through actual testing, its model inference time is approximately 0.18s, which can meet the needs of facial feature recognition in certain scenarios.
References
[1] Master's Electronic Journal Publication Information: Year: Issue 04, 2024, University of Electronic Science and Technology of China, Zhong Haoyuan, Research and Application of Lightweight Facial Expression Recognition Model for Embedded Devices
[2] Implementation of Video Expression Recognition Based on Deep Compressed Spatiotemporal Model on Edge Devices, Master's Electronic Journal Publication Information: Year: Issue 01, 2021
[3] Design and Implementation of ARM based Facial Expression Recognition System, Issue 02, 2020, Electronic