嵌入式热成像图像分类
案例介绍
(1). Introduction: This project is my first attempt to do a neural network recognizing heat-map on stm32f4. Main references are:
NNOM : https://github.com/majianjia/nnom
RE-Thread : https://www.rt-thread.org/document/site/tutorial/quick-start/introduction/introduction/
(2). instructions:
AMG883-Lenet directory is using python3 and keras to train and test amg8833 heat-map data collected by recieve.py. Trainning is in amg8833_lenet-5.py. Some function is in utils.py.
stm32f407-NNOM-AMG8833-lenet.rar file includes mainly stm32 code. Also needs some libs like CMSIS-NN(version>1.8) and Rt- Thread(version>3.0) and CMSIS(version>5.2).
This is the first version, later I'd like to upload and modify some files.
(3). Results: I trained 3 types of gestures recognition : None--0, one finger--1, two finger--2, it's about 70-80% accuracy in test, and FPS is 10.
嵌入式热成像图像分类
案例介绍
Mainly using AMG8833, RT-Thread and NNOM libs to run heat-map recognition and setup neural network on STM32F4
(1). Introduction: This project is my first attempt to do a neural network recognizing heat-map on stm32f4. Main references are:
NNOM : https://github.com/majianjia/nnom
RE-Thread : https://www.rt-thread.org/document/site/tutorial/quick-start/introduction/introduction/
(2). instructions:
AMG883-Lenet directory is using python3 and keras to train and test amg8833 heat-map data collected by recieve.py. Trainning is in amg8833_lenet-5.py. Some function is in utils.py.
stm32f407-NNOM-AMG8833-lenet.rar file includes mainly stm32 code. Also needs some libs like CMSIS-NN(version>1.8) and Rt- Thread(version>3.0) and CMSIS(version>5.2).
This is the first version, later I'd like to upload and modify some files.
(3). Results: I trained 3 types of gestures recognition : None--0, one finger--1, two finger--2, it's about 70-80% accuracy in test, and FPS is 10.
案例图片
1小时精推人才
需求方专属客服，免费梳理匹配
