SLAM Technology

Math — is the most reliable form of prediction

Demo videos

A glimpse at the future of SLAM
Dioram SLAM One powers custom VR-Headset prototype with 6 DoF inside-out racking. Demo with Point Cloud
Dioram SLAM One vs Intel T265. Pose relocalization and tracking trajectory comparison
Dioram products
SLAM one
Precise 6DoF inside-out tracking with mass market components. No expensive RGB-D cameras or external IR-markers are needed.

SLAM One is a truly platform agnostic, operating in a wide range of hardware and software ecosystems.

The middle density point cloud is a winning compromise between stability, speed and performance. Boosting of several types of descriptors is used for reliable matching leading to spectacularly robustness to volatile external conditions.

Optical flow technology is used instead of standard FEM (Feature Extraction and Matching) pipeline to significantly speed-up the re-estimation of visual landmarks during the mapping process.

Consisted of many building "blocks" SLAM One is a deep scalable and customizable. Each of the block could be fully customised to satisfy customers individual needs.
Deep track
Fully inertial tracking with IMU-array powered by Machine Learning and Reservoir Computing.

Deep Neural Networks are trained to compensate noise, drift and other types of IMU errors whilst accurately predicting pose. Mighty power of the trainable hierarchical dynamic systems allows for making a precise pose estimation even in the mid-range time intervals.

Neuro Visual SLAM subsystem integrates with SLAM One for superior performance.

Neuro Visual SLAM is end-to-end trainable system that learns to reproduce pose of a camera with respect to the world using its own internal spatio-temporal environment representation.
No hand-crafted semantics, no explicit feature extraction, just trainable mapping from a raw data to the current pose of the system.
Sensor Reactor
Low-level sensor fusion

A flexible software architecture allowing low-level mixing of raw data from a different sensors — Lidars, RGB, Time-of-Flight, Deep Cameras, IMUs, GPS and more. It expands the range of possible conditions for device usage, improves positional accuracy and minimizes errors.

The whole system is based on Hierarchical attentive model, which allows for adaptive extraction of useful information from a plain, raw input multimodal data.

SLAM solution request

Complete the form and give us information about the SLAM tracking you require.
Ask us anything about Dioram SLAM One!