Run FFmpeg in a separate process to leverage hardware acceleration

[blakeblackshear]

I was looking into support for hardware accelerated decoding and realized that OpenCV doesn’t use the hardware acceleration built into FFmpeg. I did some basic tests, and it makes a substantial difference. On the laptop where I am running frigate, decoding a 25FPS 1080p stream uses about 25% of a single CPU. Switching FFmpeg to use hardware acceleration drops that to <2%. I can read raw frames form a separate FFmpeg process like this. Once I add support for this, even low powered machines should be able to perform object detection on many cameras.

[blakeblackshear]

I am running all containers on CoreOS, and the kernel doesnt support video drivers. I will either need to build a custom CoreOS image like this or use a separate machine.

[chelming]

I too am running CoreOS and am extremely interested in where you get with this!

[blakeblackshear]

At the moment, I think it is more likely that I will run this on a separate device. I love CoreOS for the seamless automatic updates, and maintaining a custom OS build would defeat the purpose for me.

[chelming]

what about forking something like this? https://github.com/src-d/coreos-nvidia