Nimitz Pedestrian Action Recognition Dataset

A high-quality video clip dataset for pedestrian action recognition at urban intersections, collected from nighttime traffic camera footage in Honolulu, Hawaii. Each clip contains one or more pedestrians annotated via a color-coded bounding box and is sorted into one of three actions. The dataset was created by jzhang27 and last updated on Hugging Face in March 2026.

Use Cases

• Train pedestrian action recognition models based on the three annotated action classes (Crossing, Walking).
• Develop and test video-based object detection systems using the color-coded bounding box annotations.
• Analyze pedestrian behavior patterns at nighttime urban intersections based on the described source footage.

Strengths

• Annotations include color-coded bounding boxes and action labels for each pedestrian.
• Data is sourced from real-world nighttime traffic camera footage, providing a specific environmental context.
• The dataset is categorized into three distinct pedestrian actions for clear classification tasks.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect geographic and temporal bias inherent to its collection in Honolulu at night.

Provenance

Source

jzhang27 on Hugging Face.

Collection Method

Collected from nighttime traffic camera footage.

Time Range

null

Freshness

Last updated 2026-03-31 22:14:22; freshness should be verified.

Geography

Honolulu, Hawaii, USA.