Merl Shopping: 41 Video Samples for Retail Action Recognition

The Merl Shopping Dataset consists of 41 video samples curated by Voxel51 for retail-focused computer vision tasks. Updated in March 2026, the collection is specifically formatted for integration with the FiftyOne library to facilitate video analysis in shopping environments.

Use Cases

• Action recognition of shopping behaviors within video frames
• Testing object tracking algorithms on retail-specific video sequences
• Benchmarking FiftyOne's video visualization and filtering tools

Strengths

• Pre-formatted for FiftyOne library integration
• Video modality for temporal action analysis
• Specific focus on retail and shopping domain

Limitations

• Extremely small sample size of 41 records
• Lack of detailed documentation on specific label classes in the provided metadata
• Geographic bias toward US-based retail environments

Provenance

Source

Voxel51

Freshness

Last updated March 2026.

Geography

United States