SEmbAI

SEmbAI is a Carl Zeiss Foundation–funded research initiative (launched February 1, 2022) developing energy- and data-saving methods for environmental perception on embedded AI systems. The program targets real-world deployments in smart factory and smart farming scenarios, where on-device perception must be accurate, low-power, and privacy-preserving.

What SEmbAI is building

• Efficient perception at the edge: Methods to run multimodal vision and sensing on constrained hardware while minimizing data movement and energy use.
• From methods to deployments: Joint evaluations and demonstrations in industrial and agricultural environments to prove robustness outside the lab. 

Work packages (selection)

• TP4 – Data generation by simulation (Lead: Prof. Karsten Berns, RRLab): Tools and pipelines to synthesize diverse training data for embedded perception, reducing costly data collection while improving generalization.
• TP1–TP3: Physical constraints and inductive biases for wearables/embedded sensing (Lukowicz), semantic information for data- and energy-efficient recognition (Dengel), and continual learning / SNNs (Stricker).
• TP5–TP7: Dedicated hardware design (Wehn/Schöbel), smart-factory experiments (Ruskowski/Plociennik), and smart-farming experiments with explainability (Dörr).

Learn more / contact: See the official SEmbAI site for contacts and updates.