Jetson TX2 Performance

Hello,

We are working on a project of face recognition which requires the capability to stream from 4 IP camera with 1080p and 30fps. I have some confusion:

  1. Is Jetson TX2 is capable to process 4 rtsp stream with opencv and resolutio 1080p with 25-30 fps?

  2. Which version of OpenCV is supported in TX2?

  3. Also which version of Tensorflow is supported?

  4. Which sdk should be used to get the best results in facial recognition?

Thanks and Regards
Hiren Kumbhare

Hi,

1. You can check the TX1 module datasheet for information.

<b>Decode</b>
------------------------------------------------------------------------------------
H.265: Main10                                             2160p 60fps | 1080p 240fps
H.265: Main                                               2160p 60fps | 1080p 240fps
H.264: Baseline, Main, High, Stereo SEI (half-res)        2160p 60fps | 1080p 240fps
H.264: MVC Stereo (per view)                              2160p 30fps | 1080p 120fps

2. OpenCV 3.3

3. TensorFlow v1.9.0. Check this topic for more information:
https://devtalk.nvidia.com/default/topic/1038957/jetson-tx2/tensorflow-for-jetson-tx2-/

4. We don’t have an optimized example for camera+TensorFlow. If Caffe is acceptable, you can use our DeepStream SDK:
https://developer.nvidia.com/deepstream-jetson

Thanks.