Bringing Digital Twins to Life with DeepScenario and Intel® OpenVINO™

About DeepScenario 

With its AI technology, Intel® Ignite and Intel® Liftoff startup DeepScenario creates the essential link between the physical and virtual world to enable industry-specific solutions in dynamic environments. This allows companies like Bosch, Torc Robotics, or Volkswagen to create a continuous loop consisting of capturing the physical world, understanding its dynamics and using this knowledge in their custom application, such as autonomous driving. 

To achieve this, a fundamental computer vision software from DeepScenario is used. Highly advanced algorithms together with commercially available cameras turn videos of the real world into a dynamic virtual counterpart, thereby establishing the foundation for the corresponding application.

The Challenge 

More and more applications require a dynamic digital twin of the world. This allows workflows to be automated, costs to be reduced, and functional scopes to be expanded. For the development of autonomous driving, for example, digital twins of road traffic are needed to test the vehicle software in simulation.  

The transportation industry serves as another example, utilizing live digital twins to manage city traffic or track the occupancy of parking spaces. Current approaches face significant challenges, such as limited scalability due to expensive sensor setups or limited application due to simplifying assumptions.

The Solution 

This is precisely where the AI startup DeepScenario from Stuttgart comes into play. DeepScenario uses its 3D computer vision software to transform videos into dynamic virtual worlds. The solution is extremely scalable because it relies on commercially available cameras and leverages a 3D approach to ensure generalization.

For instance, DeepScenario's AI technology is used in the automotive industry. Here, the industry-leading algorithms from DeepScenario are deployed to convert videos from vehicles, traffic cameras, or drones into dynamic digital twins of road traffic. These virtual worlds are then used to train and test new vehicle software in simulation. 

DeepScenario’s software also generalizes to applications in the transportation industry. This includes, for example, precise parking lot monitoring leveraging cameras that have already been installed. The software automatically recognizes all objects in the scene in 3D with centimeter precision and tracks them over time. Unlike existing 2D-based methods, DeepScenario’s algorithms can precisely determine the position, orientation, and dimension of the objects in 3D without the need for additional sensors like LiDAR. This enables them to infer the current parking space occupancy in real time or identify potential parking violations. 

Also, completely new services can be implemented with DeepScenario’s software like Automated Valet Parking. Bosch Mobility has recently launched such a service.

Many other use cases are feasible with DeepScenario’s AI technology including intelligent traffic control or logistics applications.  

Intel Tools Support DeepScenario for Use at the Edge 

With the use of Intel® tools, we have managed to make our software executable in real time on edge servers in a very short time. This opens up a market for us with more than one billion stationary cameras and a volume of about $15 billion, with further growth expected. - Dr. Holger Banzhaf, Managing Director, DeepScenario. 

Another benefit of DeepScenario's software is the optimized interaction with Intel® hardware leveraging, e.g., the Intel® OpenVINO™ Toolkit. This open-source software accelerates the inference of AI applications on Intel® hardware, especially when developing and implementing deep learning-based AI models such as those from DeepScenario. 

In practice, this collaboration shines in a smart parking solution jointly enabled by DeepScenario and Intel® SceneScape. By leveraging the OpenVINO™ toolkit, this solution creates accurate 4D digital twins of parking lots using just existing monocular cameras. It not only overcomes the limitations of traditional methods, but also extends to various applications including smart traffic management, port monitoring, and Industry 4.0 initiatives. 

In addition, Intel® supports the development of the Carla Simulator, which is widely used at DeepScenario and in the field of autonomous driving. 

With its AI technology, DeepScenario is breaking new ground in creating dynamic virtual worlds from real-world recordings, with the support of Intel®, Intel® Ignite and the Intel® Liftoff program for AI startups. DeepScenario’s approach is extremely scalable and can be used in countless applications. The use of 3D algorithms opens up new technical possibilities that DeepScenario has already proven to be essential in the automotive and transportation industry.