Customizable, real-world human motion datasets delivered at scale — ethically collected and ready for machine learning pipelines.
Modular, task-driven human motion datasets designed for perception and control systems.
Each category can be customized by .
Single-view hand gesture footage captured using head-mounted iPhone rigs, optimized for consistent framing and repeatable motion sequences.
Everyday household actions recorded in natural environments using wearable, first-person capture for realistic activity modeling.
Human motion data recorded in operational environments using wearable capture to document task execution and workplace movement patterns.
Fine motor actions captured in controlled single-view setups, focusing on precision hand movements, manipulation, and tool use.
Sample-first datasets collected to validate task scope before scaling capture and labeling based on customer-defined requirements.
Clip length, context, task complexity defined per project.
Hours of data delivered per month via contributor network.
New task definitions deployed quickly across regions.
Standardized labeling, metadata, and predictable formats.
Core operations based in Asia. A distributed contributor network enables global-scale capture on demand.
All datasets from Intelligent Motion are delivered in a clean, structured format designed for direct integration into machine learning pipelines.
dataset/ videos/ labels.json index.csv schema.md
addons/ annotations/ keypoints/ segmentation/ splits/
qa/ spot_checks.csv label_audit.json completeness_report.md
APAC operations + efficient capture workflow to keep $/hour low.
Sample-first, then scale collection on demand with clear milestones.
Task, environment, duration, and labeling depth defined per project.
Natural settings + first-person capture for realistic model behavior.
Consent-based capture, fair compensation, and commercial usage rights.
Hand & arm motion → intent / control
Fine motor + tool-use demonstration
Daily actions in natural environments
First-person view for household tasks
Behavior cloning / policy-learning inputs
Hard-to-find scenarios captured on demand