The DNNCam is BoulderAI’s proprietary deep neural net camera system. The camera contains all elements needed to sample very large high-resolution scenes, synthesize the data through deep learning, and then store or transmit contextualized data. While it is a small device in size, the DNNCam is very large in power. In this post, we dissect the elements of the camera’s hardware.
The All-Environment Case
The DNNCam is not just a camera. The device and its components are housed in an industrial-strength case that comes with a dustproof and waterproof IP67 rating.
Even more important, the case has a passive cooling feature to ensure the hardware continues to work in the most extreme temperatures. With Boulder AI’s DNNCam, you don’t have to plan for repairing damaged cooling fans.
Deep learning requires fine details. The camera is designed with this in mind.
The motorized lens offers 10bit HDR, 4K resolution, and captures 60 frames per second making it more than capable of picking out those fine details.
The Deep Learning Processor
The processor relies on the INVIDIA TX2 module, which offers advanced graphics processing unit (GPU) capabilities and enables efficient machine learning in detail. Here’s a brief look at the INVIDIA TX2 as it relates to the DNNCam.
The NVIDIA TX2 is twice as fast and energy-efficient than its predecessor and allows users to deploy AI algorithms based on Deep Neural Networks directly at the source of data rather than moving all the data to the cloud and processing it on a remote server. This minimizes network latency that is brought about by upload and download of data. This also allows for use in areas with weak and unreliable internet connectivity.
The On-board Communication System
The on-board communications system is an important feature. With this, you are not required to have accompanying hardware. It’s all-in-one. The system can be customized to be compatible with everything from an analog cell phone system to a satellite phone to regular old WiFi.
The On-board Storage
Storage is a concern in camera systems as connection problems are common. The DNNCam comes with 32 GB onboard flash storage, which allows data to be stored if/when there’s a connection problem. Also, the storage is expandable (MicroSD), so it’s possible to deploy one of these cams to store data for days or weeks until the data can be retrieved.
You can put the DNNCam anywhere, not just anywhere you have a power outlet.
This deep learning device has a low overall power requirement (less than 15W), which makes it possible to run the camera off a solar panel and battery pack, opening the door to implementations all over the world. Satellite communication modules are available, too.
Want to learn more about the DNNCam? You can find some more details on the specs page, but we recommend you give us a call to learn how the DNNCam can be customized to suit the needs of your AI project.