OpenCV AI Kit
A Tiny Powerful Spatial AI System from the Biggest Name in Computer Vision.
Tiny Powerhouse for AI and Computer Vision
OAK consists of the OAK API software and two different types of hardware: OAK-1 and OAK-D. They are tiny artificial intelligence (AI) and computer vision (CV) powerhouses, with OAK-D providing spatial AI leveraging stereo depth in addition to the 4K/30 12MP camera that both models share. They are also both absurdly easy to use. Up and running in under 30 seconds, OAK-1 and OAK-D allow anyone to access this power: hobbyists, researchers, and professionals alike. Once you’re done tinkering, OAK’s modular, FCC/CE-approved, open-source hardware ecosystem affords direct integration into your products.
| OAK Color Camera | OAK-D Stereo Camera | |
|---|---|---|
| Shutter Type | Rolling Shutter | Sync Global Shutter |
| Image Sensor | IMX378 | OV9282 |
| Max Framerate | 60fps | 120fps |
| H.265 Framerate | 30fps | / |
| Resolution | 12MP (4056×3040 px/ 1.55um) | 1MP (1280×800 px/3um) |
| Field of View | 81° DFoV – 68.8° HFoV | 81° DFoV – 71.8° HFoV |
| Lens Size | 1/2.3 Inch | 1/2.3 Inch |
| Focus | 8cm – ∞ (AutoFocus) | 19.6cm – ∞ (FixedFocus) |
| F-number | 2.0 | 2.2 |
OAK-1 & OAK-D Features
OAK does all these AI and CV functions in arbitrary combinations thanks to the OAK API software.
Neural Inference
Object detection, image classification, semantic segmentation, pose estimation, etc.
Warp/Dewarp
Support for additional lenses for fish-eye applications
Object Tracking
Up to 20 objects with unique IDs
Apriltags
Structured navigation
H.264 and H.265 Encoding (HEVC, 1080P & 4K Video)
3.125MB/s (tiny bandwidth) for 4k video; A Pi Zero can record 4k/30fps with this.
Feature Tracking
Optical and Visual Inertial Navigation
JPEG Encoding
12MP Stills
Motion Estimation
Allows real-time background subtraction
MJPEG Encoding
For easy web streaming/etc.
Edge Detection Harris Filtering
Automatic motion-based lossless zooming (OAK-1 Specific)
- 12x lossless zoom with 720p output
- 6x lossless zoom with 1080p output
- 1.5x lossless zoom with 4k output
Stereo Depth (Including Median Filtering, OAK-D Specific)
Extended disparity and subpixel possible for wider dynamic range
3D Object Localization(OAK-D Specific)
- Monocular AI with Stereo Disparity Depth
- Stereo AI (i.e. stereo neural inference) for small object/feature support
Object tracking in 3D space (OAK-D Specific)
- 3D Trajectory in Real-Time
- Enables motion statistics in meters
Powered by Intel Movidius Myriad X
The OpenCV AI Kit namely OAK is a tiny low-power hardware edge AI computing module based on Intel Movidius Myriad-X embedded AI chip. Compared to the other AI acceleration solutions based on GPU, CPU, FPGA, or TPU, Movidius is a VPU architecture, it has 4.0 TOPS computation capacity. And it is 80 times faster for CV and AI tasks than the well-known OpenMV project that only has 0.05 TOPS based on ARM Cortex M7 microcontroller.
The OAK has the same AI chip as the Intel Neural Compute Stick 2 (NCS2) but has more powerful hardware features. OAK shipped with one 1/2.3″ Sony 12MP IMX378 capable of [email protected] H.265 video streaming, video AI pipelined processing, and two optional 1MP monochrome global shutters OV9282 cameras for depth sensing, with all 3 cameras it turns the OAK into an RGB+D camera.
| Movidius Myriad X Visual Processing Unit | Built-in to every OAK module |
| Compute Capacity | 4 Trillion Ops/sec |
| Vector Processors | 16 SHAVEs |
| Vision Accelerators | 20+ |
| Memory Bandwith | 450 GB/sec |
Movidius’ VPU natively supports Intel Openvino software deployment tool like the NCS2 does, OAK is not only Openvino compatible but even unlocked the video pipelined processing power for the neural inference engine to reveal the speed and depth capability that you have never seen on the NCS2.
With the OAK API, now it works with the OpenCV. Unlike any other Myriad X-based solution in the world, unlocking the full potential of this powerful VPU for the benefit of the entire OpenCV community. Feature tracking, hardware-level H.265 support, and 4k output are now unlocked for use by anyone.
OAK vs. Other Myriad X Solutions
| Metric | OpenCV Al Kit w/ Raspberry Pi | Other Popular CV + AI Kit |
|---|---|---|
| Picture Resolution | 12MP (4056×3040) | 1MP (1280×720） |
| Video Resolution | 4K | 720p |
| Easy Setup & Development | Yes | No |
| Efficient Data Path | Yes | No |
| Real Time | Yes | Yes |
| Low Latency | Yes | Yes |
| Embeddable | Yes | No |
| Productizble | Yes | No |
| CPU Free for User Code | Yes | No |
| CPU Utilization | Near-Zero | High |
| Hardware H.265 Support | Yes | No |
| Hardware JPEG Support | Yes | No |
| Hardware Feature Tracking Support | Yes | No |
| Power | 6W (max, including Pi) | 50 W(max) |
| Price | $199-$299 | $879.95 USD |
See What the Pros are Thinking of OAK.
Here is the understanding of OAK from the professionals!
Essentially, they’re (almost) plug-and-play eyeballs for gadgets.
Tristan Greene
The Next WebA reliable, low-cost, low-power-draw computer vision unit like this is a great boon for anyone looking to build a smart device or robot.
Devin Coldewey
TechCrunchA Piece of embedded hardware that is making it easier than ever for computer vision and deep learning practitioners to apply CV/DL to embedded devices..
Adrian Rosebrock PhD
PyImageSearchand See How the Community is Loving OAK.
We were so excited that the OpenCV AI Kit project was successfully funded in Kickstarter with a total amount of $1,358,318 US dollars. The project goal was funded in just 20 minutes.
Embedded AI for Real Products
OAK is a modular, open-source ecosystem composed of MIT-licensed hardware, software, and AI training – that allows you to embed the super-power of spatial AI plus accelerated computer vision functions into your product. OAK provides in a single, cohesive solution what would otherwise require cobbling together disparate hardware and software components.
Absurdly Easy 30-Second Setup
The OAK API is the fastest way to get started in spatial AI, especially with the OAK-1 module’s single USB-C connector which provides it with both data and power. As you can see in the video above, you can go from unboxing the hardware to running an advanced image classifier in under one minute.
Use Free Neural Nets or Create Your Own
OAK ships with neural nets covering: COVID-19 mask/no-mask detection, Age recognition, Emotion recognition, Face detection, Facial Landmark (e.g., corners of eyes, mouth, chin, etc.), General object detection (20-class), Pedestrian Detection, and Vehicle detection. This list gets added to on a regular basis.
What Does Arducam Do
Arducam proudly works with Luxonis and OpenCV for distributing the OAK-1, OAK-D, and OAK system on module (SOM) and developing more camera modules and carrier boards with a different form of factors and more optics options.
Optical Customization
OEM Compact Camera Module
Camera breakout Design
