it spent 65ms and fps=14 on recognize a apriltag,we want the recognition speed is 10-20ms,fps=50-100,
how can we speed up the recognition?use cuda or deep learning method?
Hi,
For recognition, you can train a googlenet network for your use-case.
It take around 22-23 ms for inferencing.
Read more at https://developer.nvidia.com/embedded/twodaystoademo
Can you do what you sugges on a Nvidia Jetson TX2 too, correct?
Sure.
You can find relevant sample on our Jetson_Inference sample here:
[url]https://github.com/dusty-nv/jetson-inference#classifying-images-with-imagenet[/url]
Thanks.