RetailAction: Multi-View Human-Object Interaction Dataset

This large-scale collection of multi-view video sequences focuses on the spatio-temporal localization of human-object interactions in retail settings. It provides synchronized camera perspectives to support the development of models for complex activity recognition and object manipulation tracking in real-world environments.

Use Cases

• Train spatio-temporal localization models to detect human-object interactions using multi-view video inputs
• Develop multi-camera fusion algorithms to track customer engagement with products in retail aisles
• Evaluate action recognition models on their ability to handle occlusions using synchronized camera perspectives

Strengths

• Multi-view video data for synchronized spatio-temporal localization
• Focuses on human-object interactions (HOI) in authentic retail environments
• Accepted at the ICCV 2025 Retail Vision Workshop