Spectacular AI
Spatial intelligence in cameras
Use cases
Spectacular AI solves challenging problems related to Computer Vision, Spatial AI, Sensor Fusion and SLAM. Check out the use case highlights below and contact us if they resonate with your needs. We are happy to discuss more to see if our technology can solve your problem.
Inside-out tracking
6-DoF pose tracking for AR & VR headsets, without external lighthouses. Accurate output at high frequency and low latency. Lightweight enough to run on an embedded processor in the headset.
Drones & Robots
Runs real-time on embedded devices. Ideal as an input for autonomous navigation. Supports common development boards such as Jetson Nano and Raspberry Pi. Low power consumption (well below 1W with correct acceleration).
GNSS-VIO
We can optionally fuse RTK-GPS information or other absolute position sources. Allows uninterrupted positioning during GPS outages in, e.g., tunnels or indoor spaces. Accurate orientation (yaw, pitch, roll), even with a single GPS antenna.
Mapping (beta)
Our SDK can be used for quick real-time mapping and 3D reconstructions on commodity hardware.
Product: Spectacular AI SDK
Next-generation Spatial AI engine designed for real-time embedded use cases. Produces location information by fusion of camera input and IMU, and optionally many other supported sensors.
- Designed for real-time embedded use cases.
- Independent of platform vendor services (ARKit, ARCore, etc.)
- Can leverage embedded DSPs/NPUs for acceleration
- Supports a wide range of different sensors and devices
- State-of-the-art accuracy, robustness and efficiency (see below)
| Minimum requirements | |
|---|---|
| Sensors | Monocular camera (rolling or global shutter), accelerometer & gyroscope |
| OS | Linux, Windows or Android |
| CPU | 1 core of ARM Cortex-A72 (Raspberry Pi 4) or better |
| x86-64 (Intel/AMD), Aarch64 (ARM) | |
| Optional | CV accelerator (recommended): VisionWorks (Nvidia Jetson), FastCV (Qualcomm), or Movidius (Intel) |
| Supported sensors: Stereo cameras, ToF/Lidar, RGBD (active IR), RTK-GPS |
Free
Free for non-commercial use
- Linux x86-64 (Ubuntu 18+)
- Windows
- Python SDK
- Supported devices: OAK-D
Commercial
Monthly base fee + per-device one time license fee
Free tier plus
- Aarch64 (ARM) support
- C++ SDK
- RealSense D455 & D435i
- Azure Kinect
Technology
Visual-inertial navigation done right
Whether your challenge is to track the movement of vehicles, people or autonomous things—you will need to understand the device's relative location in the surrounding 3D environment.
Our methods are learning-based and adapt to uncertainties in the observed data. We do principled probabilistic inference, where both visual and inertial data contribute as equal parties. Our unique approach is built from scratch using first-principles and implemented using state-of-the-art computer vision and machine learning methods.
Related research
O. Seiskari, P. Rantalankila, J. Kannala, J. Ylilammi, E. Rahtu, A. Solin
HybVIO: Pushing the Limits of Real-time Visual-inertial Odometry
2022 IEEE Winter Conference on Applications of Computer Vision (WACV)
A. Solin, S. Cortés, E. Rahtu, J. Kannala
PIVO: Probabilistic Inertial-visual Odometry for Occlusion-robust Navigation
2018 IEEE Winter Conference on Applications of Computer Vision (WACV)
A. Solin, S. Cortés, E. Rahtu, J. Kannala
Inertial Odometry on Handheld Smartphones
2018 21st International Conference on Information Fusion (FUSION)
S. Cortés, A. Solin, E. Rahtu, J. Kannala
ADVIO: An Authentic Dataset for Visual-Inertial Odometry
2018 European Conference on Computer Vision (ECCV)
Team
Spectacular AI is a university spin-off company from Helsinki, founded in 2021.
Our team is a unique mixture of seasoned software professionals plus established computer vision and machine learning researchers. We solve challenging problems related to Computer Vision, Spatial AI, Sensor Fusion and SLAM.
