Running Tensorflow model on jetson tx2

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hello,
i’am trying to run a deep learning model using tensorflow and openCV for emotion detection . the power mode is MaxN .the problem is that the Ram is filled before the camera is open and the frame rate is very low that there is 6 seconds delay . i used three different models. the highest number of parameters was 6 million and the lowest was 900k parameters and the performance was the same

is there any configurations to be set rather than the power to gain the maximum performance from the board ?

2

Hi,

If you don’t have too much dependency on TensorFlow, it’s recommended to use TensorRT instead.
Since TensorRT has optimized for the Jetson system, it won’t occupy too much memory as well as better performance.

Here is our benchmark result for TX2:
https://github.com/NVIDIA-AI-IOT/tf_to_trt_image_classification#models

Thanks.