How to open GPU or multiple of CPU to get real time?

Hi, this is my first time to use Jetson TX2. And I use tensorflow at python. And do image recognition with camera. But it’s always can’t be real time. And it only run one CPU, and i don’t know if it run GPU.
Then I use
sudo ~/tegrastats
to see. Then I can see my GPU(GR3D) will run to 99%. Is that mean I’m using the GPU? But it still can’t be real time. And this time I found the CPU just run one. How to open all CPU to run this program? And this way can solve this problem to get real time? Or tensorrt is the only way to that it to run fast. But I can’t find the way to use it at python.

This is the tool’s version I use on TX2:
Cuda 8.0
cudnn 7.0.1
python 3.5
tensorflow 1.3.0

Thanks all!

Hi,

This is a case-dependent problem. Could you share your use-case with us? Is a classification issue? Detection? Or segmentation?

Here is some information for your reference first.

  1. Maximize CPU/GPU
sudo ~/jetson_clocks.sh
  1. You can find device information when launching TensorFlow:
...
I tensorflow/core/common_runtime/gpu/gpu_device.cc:906] DMA: 0 
I tensorflow/core/common_runtime/gpu/gpu_device.cc:916] 0:   Y 
<b>I tensorflow/core/common_runtime/gpu/gpu_device.cc:975] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GP10B, pci bus id: 0000:00:00.0)</b>
...
  1. GR3D also can give you the information on GPU utilization.

Thanks.

Hi, AastaLLL
Thank you for you answer.

I will try your first answer later.

And the first time I tested tensorflow, it also had the code

...
I tensorflow/core/common_runtime/gpu/gpu_device.cc:906] DMA: 0 
I tensorflow/core/common_runtime/gpu/gpu_device.cc:916] 0:   Y 
I tensorflow/core/common_runtime/gpu/gpu_device.cc:975] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GP10B, pci bus id: 0000:00:00.0)
...

so it means i’m using the GPU?

I’m doing the SSD with tensorflow.

Hi,

Yes. That means you are using GPU mode.
Thanks.