专栏名称: 机器学习研究会

机器学习研究会是北京大学大数据与机器学习创新中心旗下的学生组织，旨在构建一个机器学习从事者交流的平台。除了及时分享领域资讯外，协会还会举办各种业界巨头/学术神牛讲座、学术大牛沙龙分享会、real data 创新竞赛等活动。

【推荐】用树莓派\/YOLO打造“穷人版”深度学习摄像头

机器学习研究会 · 公众号 · AI · 2017-12-20 22:33

正文

点击上方 “机器学习研究会” 可以订阅

摘要

转自：爱可可-爱生活

Poor Man's Deep Learning Camera

Build a thin client deep learning camera in Python with the Raspberry Pi, Flask, and YOLO

Installation

You'll need a few different libraries installed on the Raspberry Pi. Most notably, OpenCV 3 with Python bindings, along with Flask.

The Raspberry Pi runs the Camera-Server code, and sends back images from a webserver.

On another computer, you'll run the inference script, and it will detect whether or not there are birds in your webcam's image.

For this to run, you'll need to download and install the Darkflow weights, along with the YOLO model of your choice. Once that's installed, you should then be able to start doing inferences.

I'll add more details shortly.

Hopefully this image makes sense. We run a cheap edge computer that just sends images out of the current webcam frame, and the other computer script does the inference on that deep learning camera.