The Deep Neural Network Camera, or DNNCam, is a system capable of replacing tedious, dangerous and subjective visual tasks such as counting, sorting, and observational reporting. In short, it is artificial intelligence in a can.
It is an ultra power efficient supercomputing camera with special functions
designed for real world off-grid and rugged environments (see video below). It can operate in extreme hot, cold, and wet environments and does not rely on a fat pipe to the cloud to make actionable decisions. It contains all necessary elements to sample
enormous high resolution scenes, reduce the data through deep learning (Caffe, Tensorflow,
CuDNN), and store or transmit highly reduced, contextual information.
Built on Nvidia
The Jetson TX2 is a 256-core supercomputer the size of a business card. Per watt, it is unrivaled in its performance in image and AI tasks.
- Camera: 4k, 10-bit HDR, 60FPS
- Lens: motorized variable focus, zoom, iris, and IR filter. Low distortion, 4k resolution
- Power: 7-15W, 12-42V input for solar or industrial power
- Comm: Wifi 802.11AC, Bluetooth, CANbus, 2x USB3.0, Gb Ethernet, 2 Uarts, SPI, I2C
- Storage: 32GB onboard flash + MicroSD expandable storage
- IMU Module: Inertial measurement unit- anti-tamper
- Dimensions: Approximately 2.5″ diameter, 6″ long
- Environmental: IP67 case rating, operating temp range -20°F to 100°F
- Passive Cooling: No cooling fans to get fouled or damaged
- Made for Developers: The DNNCam has an NVIDIA TX2 GPU module onboard, widely supported by open source development community
- Perfect for Fog Computing: Looking for an AI camera with an onboard GPU for fog computing applications? You’ve found it.
Included Bonus Accessory:
- GPS Module 20Hz global position 3m
Optional Additional Accessories Coming Soon:
- Iridium satellite telemetry modem for world-wide communication
- Internal battery: Lithium polymer battery backup + solar charging
- LED lighting: 260W LED lighting driver onboard. 8-65V input, 0-1000ma output,
pulse synced with camera
The DNNCam™ does not have a cooling fan, as this would prevent the camera from being waterproof and dustproof.
Instead, it utilizes passive cooling, modeled with computational fluid dynamics.
As you can see, the engineering behind the DNNCam is extensive.
(click image to see larger version)
Current customers and application examples
The system is modular, extremely flexible and can be built to suit its environment and use case.
- Smart City:
- Trash dumping detection (Partnership in Belgium).
- Train arrival/early warning system.
- Drone and airspace clearance/detection.
- Stationary or car-mounted license plate parking enforcement.
- Underwater fish habitat and fish species ID.
- Fish ladder efficiency and quantification. (Current customers)
- Species-aware game camera.
- Contextual scene information gathering.
- Microscopy: histopathology segmentation/analysis and microbial activity tracking.
- Advertisement gaze/engagement analysis.
- Retail demographic and regular
- Recycling material handling (Current
- Department store heat map.
- Register customer flow and wait times,
automatic register open/shutdown.
- Employee activity monitoring.
- HOA License plate reading traffic flow
- Bank monitoring through human pose
estimation and behavior analysis.
- On-site table grapes sorting and disease
state classification. (Partnership in the
- Coffee sorting and value transfer in the
fields. (Current customer)
Bonus, GPS Module and Antenna are included!
About Boulder AI