search

WongKinYiu / yolor

implementation of paper - You Only Learn One Representation: Unified Network for Multiple Tasks (https://arxiv.org/abs/2105.04206)
GNU General Public License v3.0
1.99k stars 518 forks source link

Smaller models? (Low FPS) #208

Open callmesora opened 2 years ago

callmesora commented 2 years ago

Hello, I've been testing these models with Deepstream on my Jetson Xavier AGX. I can't break the 10 FPS barrier connected to a camera. Are there any smaller models I can use of YOLOR?

image

MathisFederico commented 2 years ago

I'm trying to tinker things using the yolor_tiny config given here #87 But of course there is no pre-trained weights, and for now I didn't manage to run it ...

MathisFederico commented 2 years ago

For now my only solution is to use lower image sizes

WongKinYiu commented 2 years ago

yaml style: https://github.com/WongKinYiu/yolor/issues/9#issuecomment-914887884 cfg style: https://github.com/WongKinYiu/PyTorch_YOLOv4

ashjostan commented 2 years ago

@MathisFederico I tried running the yolor_tiny config given here #87 as well.

Unsuccessful so far with the tiny_yolor.txt mentioned. I've mentioned the changes I've made for my custom dataset and still get an error - https://github.com/WongKinYiu/yolor/issues/87#issuecomment-1085409444