Oak-D-Lite Binocular Deep Artificial Intelligence Camera Machine Vision Opencv Ai Embedded

• Type: Development Board
• Trademark: OAK-D-Lite
• Origin: China
• Model NO.: Binocular Deep Artificial Intelligence Camera

Product Description

 

 

OAK-D-Lite takes the affordability idea and pushes it one step forward. By having the same SpatialAI functionality as OAK-D with a smaller weight and form factor, it gives you the opportunity to create all sorts of projects.

It is meant to be used by anyone, anywhere. The Swiss Knife of Computer Vision.

Preorder here: OAK-D-Lite Indiegogo Campaign Page or click on the image below.

Key features

• Neural Inference (object detection, image classification, semantic segmentation, pose estimation)

• Object tracking

• 3D Object Tracking (up to 20 objects with unique ID)

• 3D Object Localization

• H.264 & H.265 Encoding (HEVC, 1080P, and 4K Video) - A Pi Zero can record 4K/30 FPS video with this

• Motion Estimation (allows real-time background subtraction)

• Feature Tracking (optical, and visual inertial navigation)

• Corner Detection

• Stereo Depth (including median filtering, extended disparity and subpixel possible for wider dynamic range)

• Warp/Dewarp - support for additional lenses for fish-eye applications

• JPEG Encoding

• MJPEG Encoding for easy web streaming

Hardware Specifications

4K Camera

Stereo Pair

Vision Processor

4208x3120 - RGB Auto-Focus:

• Aspect Ratio: 1.348:1

• Focus Range: 8cm - Infinity

• Lens Size: 1/3.1 inch

• Effective Focal Length: 3.37mm

• Diagonal Field of View: 81.3 degrees

• F-Number: 2.2 +-5%

• Distortion: < 1%

• Image Sensor: Sony IMX214

Stereo Pair

300,000 depth points up to 10 meters:

• Aspect Ratio: 1.333:1

• Focus Range: 6.5cm - Infinity

• Lens Size: 1/7.5 inch

• Effective Focal Length: 1.3mm

• Diagonal Field of View: 85.6 degrees

• HFOV: 72.9 degrees

• VFOV: 57.7 degrees

• F-Number: 2.2

• Distortion: < 1.5%

• Image Sensor: OmniVision OV7251

 

Vision Processor

20 Computer Vision Processors:

• 16 "Mini-GPU" Vector Processors

• 2 Dedicated AI Processors

• 1 Dedicated 200FPS Disparity Depth Processor

• 4 Trillion Operations Per Second

 

OAK-D vs OAK-D-Lite

OAK-D-Lite is smaller, ligher and uses less power; compared to OAK-D. The main differences are:

• Mono cameras have lower resolution (640x480 instead of 1280x800)

• There is no IMU (Inertial measurement unit) on the OAK-D-Lite, as most users don't need it. IMUs are currently hard to come by, so the price increased significantly - and we wanted to lower the price point as much as possible.

• There is no power jack, as most users just used the USB-C for power delivery, which provides 900mA at 5V and is enough for most use-cases. However some functions (e.g. inference, video encoding) can lead to large current spikes, so there is a chance hosts like RPi won't be able to provide enough power. In that case, you would need to use a power adapter.

Additionally, the more technical differences are the following:

• Myriad X VPU is used directly on the PCB (instead of connecting the SoM to the baseboard - like it is done on OAK-D)

• In collaboration with ArduCam, we developed custom camera sensor connectors, which are shorter and take up less space

Fixed-focus vs Auto-focus

When ordering OAK-D-Lite, you can select whether you want Fixed-focus or Auto-focus on the color camera, with the default option being Either works for me (in case any of those options meet your needs). Note, this does not apply to the mono cameras, as they are already fixed-focus (stereo pairs fundamentally need to be fixed-focus). The main difference between those 2 modes is the minimal focus distance:

• Auto-focus can clearly focus objects as close as 8cm, while also being able to focus objects at distance.

• Fixed-focus has a much further-out minimum focus distance.

Specifically, from our tests we deduced that fixed-focus color camera was sharp after about 50cm and blurry at about 30cm. In between those distances, some pictures were blurry, some not, therefore we cannot state the exact cutoff.

You can choose which mode you want, based on the following pros and cons:

Fixed-Focus

Auto-Focus

Pros:

• Handles vibration better. If the camera is under heavy vibration, the lens won't vibrate which would lead to blurry images (drones, robots, on machines, etc.)

• Better for RGB depth alignment (as the lens does not move)

Cons:

• Objects have to be at least 30cm away to be in focus

Dimensions and Weight

• Width: 91 mm

• Height: 28 mm

• Length: 17.5 mm

• Baseline: 75 mm

• Weight: 61 g

Minimal and perceiving distances of the camera

Minimal depth perceiving distance of the camera depends on mono camera FOV, resolution, baseline and stereo depth mode, more info is available on the Stereo Depth documentation.

OAK-D-Lite has a baseline of 7.5cm and by varying the stereo depth mode, we get the following minimal depth perceiving distances:

• Min distance (480P): ~ 35cm

• Min distance with extended disparity (480P): ~ 19.6cm

• perceiving distance: ~ 19.1 meters

For more information about the maximum distance see the Stereo Depth documentation.

Power consumption

• Standby: 0.6 W

• Running depthai_demo.py: 4 W

• Max consumption: 4.5 W

Occasional power spikes of 2W may occur when running videoEncoder and/or Neural Network. OAK-D-Lite can be also powered off of USB3.

Operating temperature

We have tested OAK-D-Lite with the following command which should inflict the highest consumption:

python3 depthai_demo.py -enc color -s color left right depth

and after 90 minutes, we received the following temperatures:

• Enclosure: 56.3 °C

• MyriadX: 73.53 °C

The operating temperature range of the Myriad X VPU (die temperature) is -40 °C to 105 °C. The die temperature can get to about +35°C from the ambient temperature. Note that the stable image temperature of the camera sensor is 0 °C to +50 °C (operating temperature -30 °C to +70 °C).

3D Models

• Enclosure step model here

• PCBA step model here