So one thing that is hard to convey about megaAI is that everything happens on the device.
This isn’t some web demo, it’s not running on AWS, and it’s not even putting load on the host.
So to help show this, we made a quick video of megaAI working on one of the older/slower little Macbooks:
So in this video megaAI is running real-time object detection in parallel with h.264 30 FPS encoding (at 4mbps). You can see that despite that 4mbps rate, it looks quite good!
Note that the settings for h.264/h.265 are fully configurable.
For h.264 1080p, if you set it to variable bit rate (VBR) and max qualtiy (100%), it will go all the way up to 39mbps (as below).
So there’s plenty of headroom above this 4mbps (0.5MB/s) rate demonstrated above. But despite that low bitrate, it still looks quite clear!
Looking forward to adding more features.
For the next update we’ll post doing object detection, object tracking (so unique, trackable IDs per object), and h.264 1080p all at the same time.
Thanks for all your support!
The Luxonis Team
An FPGA-based SoM with integrated vision, audio, and motion-sensing capability
An open source stereoscopic camera based on Raspberry Pi
An embedded platform for combining Depth and AI, built around Myriad X