Using the LightBuzz SDK, you can target the largest device ecosystem possible: Android, iPhone, iPad, RealSense, OAK-D, Kinect, LiDAR, Logitech, webcams.
There are two types of cameras: RGB cameras and depth cameras. The RGB cameras are commonly found on mobile phones, tablets, and laptops. LightBuzz uses the RGB data to capture the 2D coordinates of the joints. Depth sensors perceive the world in three dimensions, allowing you to access the X, Y, and Z coordinates of a person’s joints. Based on the target platform, our SDK supports the following camera types.
|FaceTime & USB webcams|
|Microsoft Azure Kinect|
|Occipital Structure Core|
|MP4, MOV, WMV, AVI|
On Android, you can use the rear and front (selfie) cameras of your phone and capture 2D skeleton data.
iPhone & iPad
On iOS, you can capture 2D data from the rear and front (selfie) cameras found on the iPhone and iPad devices. You can capture 2D and 3D data from the iOS LiDAR camera, found in iPad Pro 2020 and later, and iPhone 12 Pro and later.
Image credit: apple.com
macOS & Windows
On macOS and Windows, you can capture 2D data from any USB camera connected to your computer. It could be the built-in laptop camera, the Apple FaceTime camera, a Logitech camera, or any other USB 2 or USB 3 webcam.
Image credit: logitech.com
When it comes to 3D body tracking on Windows, a depth camera is the way to go. Intel has released affordable depth cameras that ship worldwide. LightBuzz fully supports the Intel RealSense D400 series:
- Intel RealSense D455 (recommended)
- Intel RealSense D415
- Intel RealSense D435
- Intel RealSense D435i
- Intel RealSense L515
Image credit: intelrealsense.com
To use a RealSense camera, your computer should have a dedicated USB 3 or USB C port.
Which RealSense camera should you choose?
When it comes to choosing a RealSense camera, you need to ask what matters most for your product. Your decision should consider the field of view, depth precision, effective environment, and frame rate of the camera.
Field of view
RealSense D455 has a much bigger field of view (90×65 versus 70×43 degrees) compared to the other members of the family. That means it covers a wider portion of the physical space, and users can get good results even if they stand closer to the camera.
Due to its LiDAR technology, L515 provides smoother depth measurements in short-range (3 meters / 10 feet). However, its accuracy is decreasing significantly when moving past 3 meters / 10 feet. On the other hand, D455 is effective in longer ranges (6 meters / 20 feet). Also, it’s more convenient for small indoor spaces.
The L515 sensor is strongly affected by ambient light coming from windows, the sun, or other sources. As a result, it’s best suited for indoor scenarios. The D455, D435, D435i, and D415 cameras are not affected by ambient light, something that makes them a better choice for both indoor and outdoor scenarios.
D455, D435, D435i, and D415 can reach 60 or 90 FPS. The L515 model is limited to 30 frames per second. If a high frame rate matters (e.g., to eliminate motion blur), D455 is better suited for you.
LightBuzz recommends the Intel RealSense D455 camera for most use-case scenarios. It’s a great general-purpose camera with a great color image, outstanding depth perception, and a wide field of view.
Occipital Structure Core
Structure Core is a high-resolution depth sensor with a remarkable 50 FPS rate. Its field of view is narrower compared to RealSense D455, but the depth precision is equivalent. Powered by a single USB-C cable, it’s a great on-the-go option that will suit most use-case scenarios.
Microsoft Azure Kinect
The Azure Kinect device combines a high-end video camera and an IR depth sensor. It features a huge field of view and a high color resolution. Azure Kinect is limited to 30 frames per second, so it’s not suitable for very fast movements. The device ships to the US, UK, China, Germany, and Japan. Currently, Microsoft is selling Kinect in bulk, so you need to check with them directly.
Overall, Azure Kinect is our second choice after Intel RealSense D455.
OAK-D is manufactured by Luxonis and backed by the OpenCV team. OAK-D became an immediate success after its Kickstarter release. It is very popular among creators and tinkerers. If you are looking for an affordable 30-FPS camera to experiment with depth tracking, OAK-D is your best bet. We recommend using OAK-D to track people close to the camera (about 5 feet/1.5 meters).