I have been using Detectron2 keypoints detection for 2 months , I got some promising results but it takes upto 70% - 90% of my 8 GB Nvidia P4000 GPU. And when I incorporated with other models like yolov8 along with the detectron2 and while loading the model all together it takes up 100% of my GPU. I have gone through online still am facing same issue.
I was using R50-FPN - 1x for Detectron2 keypoints detection the lowest model.
How to make the detectron2 use less GPU ?
Please provide a support .Advice if any config or feature available to reduce this slowness.
Hey ,
I have been using Detectron2 keypoints detection for 2 months , I got some promising results but it takes upto 70% - 90% of my 8 GB Nvidia P4000 GPU. And when I incorporated with other models like yolov8 along with the detectron2 and while loading the model all together it takes up 100% of my GPU. I have gone through online still am facing same issue.
I was using R50-FPN - 1x for Detectron2 keypoints detection the lowest model.
How to make the detectron2 use less GPU ?
Please provide a support .Advice if any config or feature available to reduce this slowness.
Also share if anyone gone through the same issue.