Errors in converting yolov3 to TensorRT

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

When i test the demo in /usr/src/tensorrt/samples/python/yolov3_onnx, got errors as follows

Reading engine from file yolov3.trt
Segmentation fault (core dumped)

and i have set up virtual memory,but it doesn’t work.

2

Hi niuniu_niu1,

Do you follow the steps shared in our document?

[url]https://docs.nvidia.com/deeplearning/sdk/tensorrt-sample-support-guide/index.html#yolov3_onnx[/url]

Thanks

3

Hi,

This is a know issue.

The default workspace size it too big for Jetson platform.
Please decrease the number from 1G into 256Mb with the comment here:
[url]https://devtalk.nvidia.com/default/topic/1052153/jetson-nano/tensorrt-backend-for-onnx-on-jetson-nano/post/5349962/#5349962[/url]

Thanks.

4

it doesn’t work on jetson TX2

5

Thanks for your feedback.
We will check this and update more information with you asap.

6

Hi,

We can run the sample without any issue.
Here is some update for your reference:

1. Make sure your onnx is v1.4.1

$ pip install onnx==1.4.1

2. Update onnx_to_tensorrt.py workspace size into 256Mb

Please try it again and let us know the result.
Thanks.

7

when i use yolov3-416 i get this error
ValueError: cannot reshape array of size 43095 into shape (1,255,19,19)

also i have modified onnx_to_tensorrt.py as follows

Two-dimensional tuple with the target network’s (spatial) input resolution in HW ordered

input_resolution_yolov3_HW = (416, 416)

so how to modify output_shapes in onnx_to_tensorrt.py

Output shapes expected by the post-processor

output_shapes = [(1, 255, 19, 19), (1, 255, 38, 38), (1, 255, 76, 76)]

Thanks.

8

it works when i modify output_shapes = [(1, 255, 13, 13), (1, 255, 26, 26), (1, 255, 52, 52)]

Thanks a lot!

9

is there any demo for video?
Thanks.