search

leehaesung / YOLO-Powered_Robot_Vision

137 stars 52 forks source link

FPS #1

Open calicratis19 opened 7 years ago

calicratis19 commented 7 years ago

Hi, Whats the FPS of this implementation on Raspberry Pi? Thanks.

ashwinnair14 commented 6 years ago

@leehaesung Same question as above. what was the FPS u got ?

ghost commented 6 years ago

same question.. what kind of frame rate are you getting on a RPi3 ?

DouglasLapsley commented 6 years ago

Same question from me too please. Also, are the dependencies listed the only ones? I thought you needed tensorflow for yolo? Any guidance welcome. Thanks.

kooscode commented 6 years ago

@DouglasLapsley - Yolo runs on Darknet. There "might" be some ports of Yolo V2 to tensorflow, but there are some custom layers built into Darknet that wont translate out of the box.. Yolo V3 introduced a new custom Darknet "yolo" layer as well..

This repo is for compiling Darknet on RPI and running yolo (tiny) on RPI..

zxzhaixiang commented 6 years ago

I am actually surprised that leehaesung got yolo to work on his raspberry pi. I had tried on my Pi 3 model B+, and the only way i can get it work is to compile darknet with nnpack, and run the yolo tiny version (not the full version). With NNPACK + Yolo v3 tiny version, I got about 1FPS. See my blog for details. http://funofdiy.blogspot.com/2018/08/deep-learning-with-raspberry-pi-real.html