Jetson Nano Template Matching on GPU

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Hi,

I was looking for papers about template matching on Jetson Nano, because I need to measure the distance of the object on stereo cameras by template matching.
I’m converting cuda img to numpy img by using jetson utils. But this process is slow if the img size is big. (when img size is small fps is ~25, when img size is big fps is ~3)

These are my questions:
In paper, they made template matching on GPU and on CPU, but I did not found anything about template matching on GPU.
Does jetson utils has any code for template matching? (I cannot found)
How can I manage this matching faster? Is there a way for template matching without converting CUDA img.
I do not know how common is jetson utils image format and if there is a library that works with format.
Or there is a better way to measure distance from stereo cam? Since our image size is 1280 * 720, I think that template matching is the best one.

Thanks,

3

Hi,

Could you share which paper do you refer to with us first?

Thanks.

4

Hi,

Yes sir,
https://www.researchgate.net/publication/338476858_Performance_comparative_of_OpenCV_Template_Matching_method_on_Jetson_TX2_and_Jetson_Nano_developer_kits

Im not sure if sharing the full paper in this forum is correct. So Im sending the link.

5

Hi,

Based on the paper, they use the template matching algorithm within OpenCV.
Not sure if they use the below source. You can give it a check:

Thanks.

1 Like