Your Partner in AD/ADAS Training and Validation
Empowering Safer Roads with Smarter Training for AD/ADAS
What We Do
Our Solution
Transforming traffic video into scenarios
Video Stream Acquisition and Actor Detection
The system starts by acquiring a video stream from any camera, including CCTV. It uses an AI detection algorithm to identify all actors, such as vehicles, pedestrians, and others.
Trajectory Extraction and Enhancement:
The system then obtains trajectories for all actors using an AI tracking system. The trajectory enhancement algorithms eliminate unnecessary trajectories, such as parked cars or objects that accidentally enter the frame. Additionally, the trajectories are improved through smoothing and other algorithms.
Conversion to Geo-coordinates
The next step involves converting the trajectories into geo-coordinates. This process allows for the accurate mapping of objects to their respective positions on the real road
Scenario Extraction and Database Formation
An algorithm extracts scenarios from the trajectories, creating a database. From an hour of video, the system can extract several thousand different scenarios. For validation and quality assessment, we use the OpenDRIVE (xodr) format, overlaying the extracted trajectories on top
Scenario Format and Conversion
We use a proprietary scenario recording format that supports conversion to various popular formats, such as OSCX 2.0. This compatibility allows scenarios to be replayed in various simulation systems, for example, in Esmini
Simulation
We also support the simulation of specific road segments, for example, in Carla, where the scenarios were collected. This simulation capability is useful for the further training and validation of autonomous driving systems.
WHO WE ARE?
We are a startup focused on creating scenarios from video data captured by road surveillance cameras. Our key advantages include delivering natural scenarios with detailed data segmentation and a massive stream of scenarios sourced from hundreds to thousands of cameras worldwide
Our origins are deeply rooted in pioneering research in computer vision, robotics, and AI. United by a shared vision, we embarked on a transformative mission to unlock the potential of motion data, enabling a deeper understanding of the world’s dynamics
We specialize in AI-driven scenario segmentation, classification, clusterization, and analysis. Our advanced technology breaks down complex traffic data into distinct scenarios, categorizes them, and groups similar patterns, providing deep insights into traffic dynamics. This comprehensive approach enables precise predictions and optimisations, revolutionising traffic management and vehicle automation.
WHO WE ARE?
- Chief Executive Officer
- Andrew Ermak
- Senior Developer
- Product Manager
- QA engeener
We are a startup focused on creating scenarios from video data captured by road surveillance cameras. Our key advantages include delivering natural scenarios with detailed data segmentation and a massive stream of scenarios sourced from hundreds to thousands of cameras worldwide
Our origins are deeply rooted in pioneering research in computer vision, robotics, and AI. United by a shared vision, we embarked on a transformative mission to unlock the potential of motion data, enabling a deeper understanding of the world’s dynamics
We specialize in AI-driven scenario segmentation, classification, clusterization, and analysis. Our advanced technology breaks down complex traffic data into distinct scenarios, categorizes them, and groups similar patterns, providing deep insights into traffic dynamics. This comprehensive approach enables precise predictions and optimisations, revolutionising traffic management and vehicle automation.
Our Features
Connect your video source or use open source cameras to generate scenario data
Access unlimited amount of scenarios which can be generated on demand. Internet is full of open translations covering extensive variations of road conditions. Our team is constantly working on expanding the list of available scenarios
Extract needed scenarios based on the tags
Looking for a particular scenario? Our team constantly improves scenario analytics to be able to filter out data by tags. Current tagging system recognizes trajectory features like turns, lane changes, U-turns, tracks, buses. Future development includes more edge cases like accident or near accident situations, pedestrian crossing, traffic light violations, cut-in, overtaking, etc
Integration with common simulation platforms
Our scenarios are compatible with most of the common simulation platforms like CARLA, Esmini. Currently scenario generation is available in OpenDrive format which can be used in CARLA, Esmini plaforms for further autopilot’s training and validation. We are constantly working on expanding the list of supported platforms
Benefits
Huge dataset of real-world scenarios
Scenario classification and segmentation
Supporting different standards like OpenScenario 2.0
Additional Solutions
Partners
ASK FOR DEMO SCENARIOS AND FURTHER PARTNERSHIP
Explore extensive, pre-processed automotive-grade traffic data spanning thousands of kilometers worldwide, or upload your own videos for 3D data extraction. Obtaining precise trajectory data at your chosen locations has never been more convenient