I finally got the apollo 2.5 perception lowcost visualizer to run on my PC and the visualizer to show up.
I am running using a GTX 1080 and am confused why the visualizer is so slow
I notice the actual perception CNN is running really fast and in realtime... as I can see by rostopic echo /apollo/perception/obstacle... so the actual perception is running well.
However, the visualizer gl_camera_visualizer that shows the bounding boxes... runs very slow and is almost 3 seconds behind. If I lower the resolution of my screen it runs faster.
Any ideas why the GLFW viewer would be so slow?
I am also wondering about cudnn, however, I am not sure this is related to this problem because I believe the DNN inferencing is happening fast enough.
But I notice the perception module (running on GPU) seems to run the same and have the same performance whether I have cuDNN installed or not.
I finally got the apollo 2.5 perception lowcost visualizer to run on my PC and the visualizer to show up.
I am running using a GTX 1080 and am confused why the visualizer is so slow
I notice the actual perception CNN is running really fast and in realtime... as I can see by rostopic echo /apollo/perception/obstacle... so the actual perception is running well.
However, the visualizer gl_camera_visualizer that shows the bounding boxes... runs very slow and is almost 3 seconds behind. If I lower the resolution of my screen it runs faster.
Any ideas why the GLFW viewer would be so slow?
I am also wondering about cudnn, however, I am not sure this is related to this problem because I believe the DNN inferencing is happening fast enough.
But I notice the perception module (running on GPU) seems to run the same and have the same performance whether I have cuDNN installed or not.