Lorem ipsum dolor sit amet, consectetur adicing elit ut ullamcorper. leo, eget euismod orci. Cum sociis natoque penati bus et magnis dis.Proin gravida nibh vel velit auctor aliquet. Leo, eget euismod orci. Cum sociis natoque penati bus et magnis dis.Proin gravida nibh vel velit auctor aliquet.

  /  Project   /  Blog: Launching YOLObot!

Blog: Launching YOLObot!

Go to the profile of MoFlo

Smart Cameras are a combination of three emerging and powerful trends. Our YOLObot app is one manifestation of these new capabilities.

YOLObot app screenshot

Sensors, AI and embedded processing.

The technical ‘dividend’ afforded by modern smartphones is immense and is still yielding amazing developments at a high cadence as the largest silicon and software platforms from Apple, Sony, Samsung, Google, Qualcomm, Panasonic, MediaTek, and others continue their push towards what was once nostalgically termed ubiquitous computing.

Smart Cameras benefit specifically from CMOS based CCDs in the 12- and now 24-megapixel range. The technical trend is to add on-die processing capabilities for standard camera features like autofocus and compression, but, uniquely, CCD vendors are adding programmable object detection. While it’s not clear whether these features will be leveraged by smartphone manufacturers, their availability means that potentially all cameras will eventually become smart cameras.

Advances in artificial intelligence arguably developed independent of the smartphone technology roadmap, but just as the leading platforms incorporated the latest communication and computational processors, all of the major smartphone technical platforms now include “neural processors” to accelerate AI/ML workloads. We expect this trend will continue with the latest embedded AI processors allowing for programmable or ‘dynamic’ neural network architectures. Whereas current processors are essentially a fixed or limited architecture, the trend is clearly to allow LLVM-based compiled networks, which means we are able to close the loop between state-of-the-art network architectures and user-land apps.

Finally, all of these advances are notably embedded. This term encompasses so much of what we now take for granted but the ability to have a battery powered device which is running smart camera workloads with 20MP class sensors and 4K display — in your pocket — is stunning. The trend here is clearly to enable more embedded, low-power solutions; both within the smartphone platforms but also as separate wearable and mounted devices.

We’re glad to release the YOLObot app as a small step along the way towards the smart camera future.

Source: Artificial Intelligence on Medium

(Visited 2 times, 1 visits today)
Post a Comment