Estimating Performance for Image Processing in Jetson AGX

Hi All,

Is there a thumb rule to estimate performance in Jetson AGX for some of the API’s ?
We are currently running application in 1080 ti. Want to have a ballpark estimate in AGX so that we can decide to move to AGX.

Please share which APIs you are interested in. We have some profiling results/comparison for running Deep Learning Inference. Need to check if there is existing data for the APIs.

Image resolution that we are using is 1000x1000.
The APIs that we are interested in are OpenCV SURF & BF Matcher.

We don’t have data of using the OpenCV functions. Would need other user to share experience.

We have OpenCV package in SDKManager installation, but it does not enable CUDA filters. If you need the functions, you would need to build/install OpenCV manually.

We are using OpenCV’s cuda enabled APIs.
SURF_CUDA from opencv2/xfeatures2d/cuda.hpp & BFMatcher from opencv2/cudafeatures2d.hpp.