DanielChai
Posts: 15
Joined: Mon Jan 21, 2013 5:29 pm

"Real-time" object detection on Raspberry Pi

Wed Aug 29, 2018 3:18 am

Tried the instruction in https://github.com/shizukachan/darknet-nnpack and wrote some Python wrapper, I managed to run "real-time" object detection using Raspberry Pi camera + Yolo object detection algorithm. The frame rate is about 1FPS on RPI v3 model B+. Here is the results

Image

Check my blog:)
http://funofdiy.blogspot.com/2018/08/de ... -real.html

I think 1FPS could really enable some very fun project, like robotic cars, drones, surveillance cameras, etc. Maybe not something super fancy. But, hey, it is only a $35 device!

ChicagoBob123
Posts: 79
Joined: Fri Aug 30, 2013 1:48 pm

Re: "Real-time" object detection on Raspberry Pi

Fri Aug 31, 2018 2:01 pm

You need to investigate the Movidius Compute stick. Costs a bit but it will further your journey in Deep Learning

mic_s
Posts: 92
Joined: Sun Oct 26, 2014 4:15 pm

Re: "Real-time" object detection on Raspberry Pi

Fri Aug 31, 2018 5:27 pm

You dosn't need to invest in a Movidius Compute Stick. The very well dokumented GPU Processing Units in your extrem-low-cost Rasp-Zero are perfectly adequate for a journey in deep learning / object recognition :

details:
https://youtu.be/zRaQS5MA2fU
(no additional hardware, besides the cam and the sd is required)
.

victagayun
Posts: 127
Joined: Sun Apr 15, 2012 4:01 pm
Contact: Website

Re: "Real-time" object detection on Raspberry Pi

Mon Nov 05, 2018 9:07 am

Could you reduce the resolution to increase the framerate?
openMV and pixycam2 are using only 640 x 480.

victagayun
Posts: 127
Joined: Sun Apr 15, 2012 4:01 pm
Contact: Website

Re: "Real-time" object detection on Raspberry Pi

Wed Nov 07, 2018 6:13 am

@ Install qmkl

Code: Select all

sudo apt-get install flex
git clone https://github.com/Terminus-IMRC/qpu-assembler2
cd qpu-assembler2
make
I got this error

Code: Select all

[email protected]:~/qmkl $ make
make: *** No targets specified and no makefile found.  Stop.

Return to “General discussion”