How to speed up ssd_mobilenet_v1_fpn with tensorRT?

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

I have try some of ssd models in Tensorflow object detection API.But “fpn” is 640x640,I don’t know how to speed it.

2

Hi,

Have you tried to accelerate it with TensorFlow-TRT?
Here is a tutorial for your reference:
[url]https://github.com/NVIDIA-AI-IOT/tf_trt_models[/url]

Thanks.

3

I have tried this github, but it doesn’t speed up ‘ssd_mobilenet_v1_fpn’.I think that TensorRT doesn’t seem to support FPN.

4

Hi,

Have you tried to convert the model into pure TensorRT PLAN?
[url]https://github.com/NVIDIA-AI-IOT/tf_to_trt_image_classification[/url]

If the whole layer are supported, it should give you some speed up.
Thanks.

5

Any update?