The perception and understanding of dynamic environments rely heavily on visual data, which has traditionally been captured using RGB cameras. However, alternative visual domains —such as thermal imaging, event-based cameras, LiDAR, hyperspectral sensors, and radar—offer new opportunities and challenges in detection, tracking, and forecasting tasks. These unconventional modalities provide valuable advantages in scenarios where standard RGB-based approaches struggle, such as low-light conditions, adverse weather, and high-speed motion analysis.

The complexity and diversity of these data sources require the development of specialized processing techniques, often involving novel Deep Learning methodologies. Event-based cameras, for instance, capture asynchronous intensity changes at a microsecond scale, enabling ultra-low-latency motion tracking. Thermal imaging can enhance pedestrian detection in nighttime conditions, while LiDAR and radar enable accurate distance estimation for autonomous systems. Additionally, hyperspectral imaging can extract material properties beyond what standard cameras perceive, contributing to applications in medical diagnostics, agriculture, and security.

Despite their advantages, alternative visual sensors introduce unique challenges, including domain adaptation, sensor fusion, and real-time processing constraints. As industries increasingly integrate these technologies into safety-critical applications—such as autonomous driving, industrial inspection, and environmental monitoring—there is a growing demand for efficient and robust models capable of handling multi-modal data streams.

This workshop aims to foster collaboration between researchers and industry practitioners, encouraging the exchange of ideas on novel sensing modalities, algorithmic advancements, and real-world applications. Furthermore, the ethical and privacy implications of these alternative visual domains will be explored, addressing concerns related to surveillance, biometric recognition, and data security.