Optimize caffemodel to run faster on Jetson TX2

I am trying to speed up inference time of OpenPose on the the Jetson TX2.

OpenPose - https://github.com/CMU-Perceptual-Computing-Lab/openpose

Built OpenPose and Caffe using the following script - https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/doc/installation_jetson_tx2_jetpack3.3.md

Unfortunately, inference time is ~1.5 FPS.

I am trying to use TensorRT, for example, to convert the model to FP16 to leverage NVIDIA Tensor Cores. Though, it is said that TensorRT python API is not supported on Jetson platform due to pyCUDA, so it’s not possible to write a python script that takes a ‘.caffemodel’ file and optimize it.

Some more specs:
TensorRT version: 4.0.2.0-1+cuda9.0
L4t version: 8.2
Ubuntu 16.04

Are there any tools/scripts available that can optimize a caffe model given it’s ‘.caffemodel’ and ‘.ptototxt’ to run faster on the Jetson TX2?

Hi,

Please try our latest JetPack release:
https://developer.nvidia.com/embedded/jetpack

TensorRT python API is supported from JetPck4.2.
Thanks.

Hi AastaLLL,
The latest JetPack Version that is available to download is “JetPack 4.1.1 Developer Preview”.
Could you please point me to a download link?

Thanks

Hi alex.g,

Please use the new NVIDIA SDK Manager to install JetPack4.2, be sure to follow the instructions.
Installer
Instructions to Download and Run SDK Manager
Instructions to Install Jetson Software with SDK Manager

Thanks