Real-world scene demonstrating multiple object detection

So, I couldn't share what in particular I am working on - but I got permission to share these scenes with those interested, and then it occured to me that would mean most of you here too.

This demonstrates the object detection which will be available in realtime, at the edge, with the extremely low powered Akida. Although, the desktop device struggled to do this and record - Akida would not struggle to detect, though. This is important to remember.

This is an example with models created very quickly through one shot learning. I excluded the model learning phase as @uiux has demonstrated it many times over now.

What's new is single and many object detection all in one scene, often detected more than once per model in a scene, and a more real-world scenario - such as, walking through a local furniture store.

Please excuse the quality and frame rate. It was screen captured whilst detecting in real time, and then converted to GIF. This is something a standard desktop computer struggles to do, but something Akida would find very easy at a high frame rate (the object detection)... which slowed the emulating PC down when doing itself... this demonstrates the marvel of what Akida really will be capable of, plus it will be done at the edge, in low power. Amazing.

Again, this is demonstrating through emulation, not the Akida chip itself. It is a simulation just like @uiux has demoed previously.

When you open these, don't forget to click the play GIF button.

Demo 1: A single type of chair which is learned and then detected many times throughout the scene. Click to open.

Demo 2: Various small objects which are learned and then detected throughout the scene. Click to open.