CANoe4AI
With CANoe4AI you test and analyze AI software components using AI-specific test methods. Already at an early stage in virtual SIL environments without real hardware. The tool supports data- and scenario-driven testing and can be flexibly integrated into existing development environments such as CI/CT workflows. This allows you to improve software quality and detect errors at an early stage.
In combination with CANoe, you get a test environment from a single source for evaluating entire systems including networks and I/O.
Challenges of Testing AI Algorithms
In more and more devices – e.g. ECUs in the automotive environment or medical controllers – AI algorithms are taking over the processing of large amounts of data. Many of these devices work in realtime. However, many AI models, especially complex ones, are not deterministic and their decisions are not easily understandable for humans. This makes troubleshooting, testability and trust in the reliability of the systems very challenging.
Vector has decades of experience in testing embedded software and complete networked systems. With CANoe4AI, we are expanding our product range to include the testing of embedded AI algorithms. CANoe4AI is currently in the pre-release phase and has been extensively tested with generic use cases. Before the release, we would also like to test specific use cases with the tool.
CANoe4AI – Key Features
- Flexible Integration
Use it as a web app or Python library to fit into complex AI setups and CI/CD pipelines. - Smart Metric Analysis
Breaks down performance metrics to identify underperforming samples. - Understandable AI Decisions
Uses explainable AI methods to show what influenced the AI’s decision—making errors easier to spot.
- Similarity Search And Dataset Generation
Finds similar samples using image or text queries to build better datasets. - Broad Support for AI Tasks and Formats
Supports classification, detection, and regression tasks for images, videos, and time series in PyTorch, TensorFlow, ONNX and more. - Full Vector Toolchain Integration
Seamlessly connects with other Vector tools to create end-to-end AI workflows.
