Yolov2 Object detection performance comparatively too slow on jetson tx2

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi,

We are developed custom code on top of the yolov2 darkflow in python. When we are trying integrate on to jetson, we didint as much performance on jetson tx2, where i will add more commands to improve the performance eventhough i wont able to achive higher fps. Here are there;

sudo nvpmodel -m 0
sudo ~/jetson_clocks.sh

is there anything i missed?

Im refering this github; GitHub - thtrieu/darkflow: Translate darknet to tensorflow. Load trained weights, retrain/fine-tune using tensorflow, export constant graph def to mobile devices. I got more examples only present cpp the sameway when i try to implement tensorrt optimizer also i won’t able to achive. Please help me, or suggest me for my usecase. Thanks in advance.

2

Hi,

To have the best performance, it’s recommended to use our TensorRT for the deep learning use case.
You can follow this sample to check if anything you can improve for the performance.
[url]https://github.com/vat-nvidia/deepstream-plugins[/url]

Thanks.

3

Thanks for the response, but it has implemented on cpp, is there any python based yolov2 tensorrt examples available? im stuck on this.

4

Hi,

Sorry that we only have C++ version for YOLO tutorial.
It’s recommended to give it a try : )