Inference with webcam crashes after ~50 Frames

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Description

Inference with webcam crashes after ~50 Frames → TensorRT execution time increases from 70ms to ~2s before

grafik
grafik1458×1017 35.4 KB

I assume a systematic problem with the while loop

Environment

JetPack Version: 4.4
TensorFlow Version: 1.15.3

Relevant Files

inference.py (3.9 KB) frozen_out_dec__batch.onnx (2.3 MB)

2

Hi, Request you to share your model and script, so that we can help you better.

Alternatively, you can try running your model with trtexec command.
https://github.com/NVIDIA/TensorRT/tree/master/samples/opensource/trtexec

Thanks!

3

Script and model are there:
https://forums.developer.nvidia.com/uploads/short-url/iMMMSLzQDIMJG6e07vV4duySrA8.py
https://forums.developer.nvidia.com/uploads/short-url/d607NNL1yylDBcVPR6oYmfDEf9K.onnx

I assume a problem with either OpenCV Framebuffer or CUDA Graficbuffer. Maybe you can have a look into the source code and give me a hint.

Thanks and have nice day!

4

Hi @toni.sedlmeier,

Problem is you are creating a new stream and context in the inference loop.
That should be hoisted out and passed in as an argument to the function IMO.

Please try fixing it.

Thank you.

5

Thanks you were absolutly right.
Which performance (fps etc.) can be reached with a non-custom trained SSD Mobilenet V2 and TensorRT API on Jetson Nano?

6

Hi @toni.sedlmeier,

Hope this will help you.
https://developer.nvidia.com/embedded/jetson-nano-dl-inference-benchmarks

Thank you.