Autonomous AI Solutions for the Next Generation of Vehicles and Mobility
With a multitude of key features, Cartica’s visual intelligence platform provides the solution for many existing challenges in the automotive industry spanning far beyond the capabilities of deep learning.
Self Learning artificial intelligence
As the system is capable of unsupervised learning, it does not require an annotated training set.
This unsupervised AI autonomously identifies commonalities among raw data to cluster footage together by concept and event.
The insights gathered from this process allow for full edge case coverage as well as a complete understanding of detailed scenarios.
Minimal Power Consumption
Cartica’s low power consumption AI crosses existing technological barriers of safety per watt for ADAS and AV.
Power consumption 10 times lower than the best available solution in the market.
The flat architecture runs on existing hardware accelerators to minimize overall compute power
Robust perception in all conditions
High accuracy visual perception in all weather and lighting conditions including fog, rain, snow, night, and more.
Programmatic coverage of longtail and edge cases
Robust concept coverage with recognition of thousands of object types (fine grain and broad)
Predictive AI understands TRavel paths
The predictive driving module provides the platform with a clear real-time risk management analysis of the environmental model, allowing it to adapt concept prediction to the driving scenario.
Knowledge Sharing (V2V, V2i, V2X)
Signature-based technology enables knowledge sharing between vehicles in an efficient way
Automatically updates local vehicles and system
The Cartica platform is able to fuse multiple sensor inputs into a single lightweight representation space. This allows for a full understanding of the environment in harsh scenarios (torrential downpour, heavy fog, etc.)
This signature fusion leverages the expressive benefits of any added sensor without the limitations of rule based fusion.
Localization & Mapping
Utilizing any visual cue as a landmark, Autonomous AI maps visual features to high-dimensional, linear signatures for localization and easy mapping updates.
A hybrid cloud/local architecture holds a small subset of signatures in the car
Previously generated signatures are reused and matched with the local cached mapping signatures. Location is determined based upon signature matching.
Unmatched signatures are sent to the cloud for processing to update the global database