Open Source Linux Camera with AI.

Description

We believe the Maker community, alongside the Open Source movement, has driven innovation into the electronics industry. The Arduino community has been revolutionary to the electronic space, allowing beginners to easily develop new and fresh ideas with relative ease. Not only by selling readily available kits, but also by Open Sourcing their schematics and PCB files.
This revolution has not yet reached the Embedded Linux world.

While Raspberry Pi has done a wonderful job selling kits, they don't provide schematics and PCB files for their boards. That's why we fully Open Sourced our product.

Von VisionAI a powerful and versatile development platform running on Linux, with an onboard NPU and 4K camera that enables you to quickly design and deploy Edge AI applications. We offer you an easy to use platform to bring your most wild ideas to life!

Subscribe to our Crowdsupply pre-campaign if you like it!
https://www.crowdsupply.com/punto-maximo-llc/von-visionai

Details

Being fully open source, Von VisionAI features a completely integrated hardware design that is not locked into a third-party System-on-Module format. It features a 7-inch 1024x600 90 Hz touchscreen, a 48 MP Sony camera, 4 GB of LPDDR4X SDRAM, 2.4 GHz Wi-Fi connectivity, Gigabit Ethernet, and USB Type-C.

Kicad Board File.

We're also including an ESP32 on our board so you can easily interface the complex and advanced Linux CPU, with the simple and well-known ESP32, so that you can interface with the outside world as easily as you are used to! Any specific object detection or data coming in or out the powerful microprocessor can be easily relayed to the ESP32, from where you can a lot more easily control motors, relay information wirelessly, turn on or off any other device, and anything you can imagine!.

While we initially developed just a CCTV camera, we also developed a portable version with a touchscreen as well for mobile applications! The poweful Mali GPU inside will offload the graphics processing off the CPU.

CCTV version of Von VisionAI powered through PoE

Portable Version. Enclosure not final.

Thanks to its integrated design, Von VisionAI is a great solution for use cases such as:

Deploy outdoors for Edge AI recognition scenarios, while on battery power
Install with standard PoE splitters far from power outlets
Apply real-time machine learning to a high-resolution video feed
Stream H.264 or H.265 video through RTSP or WebRTC
Analyze multiple Ethernet cameras' worth of data on one device

Currently we benchmark slightly over 30fps doing AI inference on 1920x1080 video streams. We offload AI to the NPU cores, and we offload rendering to the GPU, leaving the main CPU around 80% free to do any other task. Everything (Schematic + PCB + Firmware) is being released under MIT license.

Real Time AI test recorded from a Von VisionAI prototype

Success Story: We wanted to save lives and reduce accidents on the streets. So we're currently doing test deployments of Von VisionAI (CCTV version) to reduce traffic infractions and violations on real time. We will disclose more about this in due time. This is just one of the infinite use cases of our platform.

Feel free to use them for your own products and ideas! Subscribe to our Crowdsupply pre-campaign if you like it!

https://www.crowdsupply.com/punto-maximo-llc/von-visionai

Also we are opening up our Discourse forums, feel free to join and say hello:

https://forum.vonvisionai.com

Files

Board.kicad_sch

kicad_sch - 4.12 MB - 07/22/2024 at 00:02

Download

Board.pdf

Adobe Portable Document Format - 4.76 MB - 07/22/2024 at 00:02

Preview

Board.kicad_pro

kicad_pro - 24.73 kB - 07/22/2024 at 00:02

Download

Components

1 × Rockchip RK3566 quad-core ARM Cortex-A55 @ 1.8 GHz

1 × Mali GPU supporting Open GL ES 1.1/2.0/3.2, OpenCL 2.0, and Vulkan 1.1

1 × H.264 and H.265 hardware encoding support up to 1080p@60 fps or 4K@20 fps. Decoding support up to 4K@60 fps.

1 × 1 TOPS NPU to offload artificial intelligence tasks from the CPU

1 × 4 GB LPDDR4X system memory

Project Logs

Collapse

Teardown 2024

avi448 • 07/26/2024 at 13:45 • 0 comments

Back in June I attended Teardown 2024 in Portland, Oregon. It is a yearly event by and for hackers and makers. It was definitely a great experience and the first time ever I was surrounded by such a gigantic group of like-minded people. Even during my college years, I was almost alone on the Maker world.

At Teardown 2024, I gave a short talk on some of the development history of this project, I tried to quickly explain some of the most important design decisions and difficulties during development, like why choose Buildroot OS? I also did a Live Demo of the Portable version at the 21:05 minute mark! I wanted to showcase the zero latency, high framerate nature of the project, even while doing AI inference on real time.

That talk was just uploaded on Youtube and you can watch it down below!

Sneak Peek!

We're currently waiting for our new enclosures. We're trying both CNC for a headless enclosure, and SLS for the portable enclosure. We believe the headless enclosure will prove great for use cases where portability is not required. We expect to be able to show them next week, for the moment being, take a look at the 3D model of the new CNC headless enclosure!

Thank you!

If you like what you are seeing, feel free to join into our Discourse forums. Currently, no one else is registered yet, but the development team will be able to reply personally to you at the earliest! We want to build it as a safe and fun way to discuss anything, ranging from possible implementation or ideas, to technical questions. We hope to see you soon!

https://forum.vonvisionai.com/

Schematics Uploaded!

avi448 • 07/22/2024 at 00:05 • 0 comments

Schematics are now up in both this project and on Github. PCB files will be uploaded later on. We will be releasing everything (PCB + Schematic) in source format under MIT license. Feel free to check it out the schematics and have a look at them!

A little something for fun

avi448 • 07/20/2024 at 17:03 • 0 comments

Just a little something for fun: It does run Doom. Chocolate Doom to be precise. The GPU is a bit beefy so I'm sure it can run far more demanding games. As for the main purpose of the project: AI Camera. We use the GPU to offload frame rendering, scale frame operations and bounding boxes drawings off the CPU.

Discussions

Open Source Linux Camera with AI.

Description

Details