Low video resolution output with video-viewer and inference

[jdeglint]

I am working through through the Jetson AI Fundamentals - S3E2.

I downloaded the jellyfish video and ran:

./imagenet.py images/jellyfish.mkv images/test/jellyfish_out.mkv

Here is the first input frame (from a screenshot):

Here is the corresponding first output frame (from a screenshot):

Why is the output video pixelated and low resolution?

I notice this same behavior when I run headless with SSH using rtp:

video-viewer /dev/video0 rtp://MY.IP.ADDRESS:PORT.NUMBER

Please help.

[DaneLLL]

Hi,
The setting to encoder would need to adjusted to get better video quality. Please check the code in
jetson-utils/gstEncoder.cpp at 833fc7998e34d852672277730a11aeed90024959 · dusty-nv/jetson-utils · GitHub

[dusty_nv]

You can set the encoder bitrate from the command-line with the --bitrate argument - see here for more info:

• jetson-inference/aux-streaming.md at master · dusty-nv/jetson-inference · GitHub

The output resolution should be the same as the input. Also, if you are running in the container, you can try building the project from source and running outside the container. Sometimes the encoding quality is lower inside the container (under further investigation)