Datasets
VideoRobotics

Long-Context CCTV

Continuous CCTV footage at extended duration, capturing the slow buildup, the event itself, and the aftermath in single uninterrupted timelines. The long sequences give long-context video understanding, anomaly detection, and temporal reasoning models the multi-hour signal needed to learn behavioral baselines and rare-event precursors.

Hours31.6K+
Scenes10+
Duration1-24+ hrs

Training Use Cases

Long-context video understanding
Anomaly and rare-event detection
Long-horizon video summarization
World models for long-duration video generation
Key Highlights
10+ scene contexts covered including traffic, retail, public space, transit, residential, commercial, and industrial
Continuous timelines spanning 1 to 24+ hours per recording, with no pre-segmentation
Fixed-position camera framing held consistent across the full duration of each recording
Behavioral baseline footage kept alongside the rare events, not stripped out

Metadata Fields

durationLength of clip in HH:MM:SS
resolutionPixel dimensions (e.g., 1920x1080, 3840x2160)
frame_rateFrames per second (e.g., 24, 30, 60, 120)
contains_audioWhether the clip carries an audio track (boolean)
primary_categoryDominant content category assigned to a video
stylefixed_position | wide_angle | dome
scene_contexttraffic | retail | public_space | transit | residential | commercial | industrial
event_presentWhether a notable event occurs in the timeline (boolean)
baseline_segmentWhether the clip includes pre-event behavioral baseline (boolean)