search

ProjectNeura / LEADS

Enable your racing car with powerful, data-driven instrumentation, control, and analysis systems, all wrapped up in a gorgeous look.
https://leads.projectneura.org
Apache License 2.0
257 stars 54 forks source link

Test: OpenCV Performance on Raspberry Pi #254

Closed ATATC closed 2 months ago

ATATC commented 5 months ago

Huge progress has been made in regard to video support in the past few pull requests. However, concerns about the performance of the Raspberry Pi do exist as the testing platform, Apply M3 Pro, is only able to provide a video feed at almost 30fps. Therefore, it is likely that OpenCV runs as slowly as GStreamer does on the Raspberry Pi. This poor performance wouldn't affect the dashboard anyway though.

Ideally, this is what we have accomplished so far:

image image

The monitor is just an additional widget. You can still switch back to other widgets. It now supports up to 4 cameras, of which 1 is enabled in this case, which are front view, left view, right view, and rear view respectively.

ATATC commented 3 months ago

The performance of the raspberry barely supports OpenCV. The bottleneck seems to be encoding and decoding.

ATATC commented 3 months ago

Trying on a virtual machine to see the minimum requirement.

ATATC commented 3 months ago

The bottleneck is Pillow.

ATATC commented 3 months ago

The program consumes 9% of an AMD Ryzen 9 5900X 12-Core Processor.

ATATC commented 3 months ago

If use OpenCV instead, it consumes 10%.

ATATC commented 2 months ago

@qmascarenhas Is it possible to schedule this test for Monday during the second period or after school?

qmascarenhas commented 2 months ago

yes - we could do that.

ATATC commented 2 months ago

@qmascarenhas That period just got occupied. What about Tuesday's last period?

ATATC commented 2 months ago

Raspberry Pi fails to run devices_visual.py.