Preview ready-to-license first-person video recorded with wearable cameras. Built for robotics, embodied AI and computer vision teams that want to evaluate real-world data before committing to a custom collection project.
Household
Training bimanual grasp and re-grasp policies on deformable objects. Captures the stabilising hand, mid-task re-grips, and natural recovery from slip.
Household
Training whole-body mobile manipulation across a lived-in space — reaching, wiping, and clearing surfaces amid natural clutter.
Household
Training tabletop tidying — grasping mixed small objects, wiping the surface, and returning items to place.
Household
Training bimanual manipulation of highly deformable fabric — flattening, aligning edges, and forming repeatable folds.
Household
Training large-fabric handling over a wide workspace — spreading, smoothing, and tucking sheets that drape and buckle.
Household
Training load-and-unload manipulation at an appliance — transferring limp garments into and out of a drum and operating controls.
Household
Training load-and-unload manipulation at an appliance — transferring limp garments into and out of a drum and operating controls.
Household
Training contact-rich cooking manipulation — cracking, pouring, and pan handling with a tool under changing material state.
Household
Training dexterous deformable-dough manipulation — rolling, shaping, flipping, and pan transfer of soft, sticky material.
Household
Training repetitive bimanual food assembly — stacking, portioning, and ordered layering across many identical units.
Household
Training multi-step food prep — measuring, rinsing, pouring, and pot handling through a sequenced cooking process.
Household
Training long-horizon kitchen manipulation — combining tool-use, pouring, stirring, and heat-side handling across a full recipe.
Household
Training wet deformable-fabric manipulation — scrubbing, wringing, and rinsing limp cloth under changing weight and grip.
Household
Training repetitive bimanual food assembly — stacking, portioning, and ordered layering across many identical units.
Household
Training whole-body pick-and-place across a cluttered space — sorting mixed objects and returning them to defined locations.
Household
Training constrained-space wiping and reaching — cleaning narrow, partially occluded areas around fixed furniture.
Household
Training vertical-surface tool-use — sustained wiping over a large flat plane with consistent coverage and pressure.
Household
Training long-horizon deformable manipulation through multi-step paper folding. Captures ordered fold sequences, crease reinforcement, and mid-task re-alignment.
Household
Training precise placement of rigid objects into tight slots — grasping, orienting, and inserting books against neighbours.
Household
Training placement of varied rigid objects into structured holders — orientation and slotting under spatial constraint.
Household
Training bimanual assembly of soft goods — stuffing, shaping, and arranging compliant filled material.
Household
Training controlled pouring and reach — guiding liquid to targets while managing container weight and spill.
Household
Training wet-object bimanual manipulation — grasping slippery items, scrubbing with a tool, rinsing, and controlled placement.
Household
Training wet-object bimanual manipulation — grasping slippery items, scrubbing with a tool, rinsing, and controlled placement.
Household
Training wet-object bimanual manipulation — grasping slippery items, scrubbing with a tool, rinsing, and controlled placement.
Fabrication
Training two-hand operation of a high-force powered tool against rigid material. Captures the firm dual-grip, controlled feed into the cut, and steady reaction-force management as the grinder bites — high-load tool-use that free-space manipulation data can't provide.
Fabrication
Training precise tool guidance under an active high-energy process. Captures the hand striking and maintaining an arc — controlled electrode angle, standoff distance, and steady travel along the joint — fine positioning under extreme conditions that no simulated data reproduces.
Agriculture
Training bimanual tool-use on compliant vegetation in unstructured terrain. Captures the coordinated split of roles — one hand bunching and tensioning grass, the other drawing a curved blade through it — under variable resistance that clean tabletop tool data can't reproduce.
Agriculture
Training bimanual selective picking on live plants in unstructured field conditions. Captures one hand steadying the stem while the other locates and detaches ripe produce without damaging the plant — delicate targeted harvesting that clean tabletop data can't reproduce
Industrial
Training precision bimanual tool-use for electronics assembly. Captures the steadying hand fixturing the board while the tool hand places the iron with sub-millimetre accuracy — controlled contact, dwell, and withdrawal that vision-only policies can't learn.
Industrial
Training whole-body bimanual handling of large rigid parts in a working facility. Captures lifting, orienting, and placing heavy sheet-metal panels — including two-person coordinated moves — behaviours that tabletop manipulation data can't provide.
Industrial
Training bimanual handling of large flexible-framed tools in a production setting. Captures lifting, positioning, and registering a screen-print frame — two-hand control of an awkward, semi-rigid object against a fixed print bed that tabletop data can't reproduce.
Industrial
Training bimanual guidance of deformable fabric through a powered machine. Captures both hands feeding, aligning, and tensioning cloth under the needle in real time — continuous fine control of limp material that tabletop pick-and-place data can't provide.
Industrial
Training bimanual pressing of deformable material with a handheld tool. Captures one hand smoothing and positioning cloth while the other guides the iron with controlled pressure and path — coordinated flatten-and-press behaviour on limp fabric that rigid manipulation data lacks.
Industrial
Training fine bimanual manipulation of soft fabric with a small tool. Captures both hands working in close coordination — one holding and tensioning cloth, the other trimming stray threads — precise close-range work on deformable material that rigid-object policies never encounter.
Industrial
Training bimanual pick-and-pack from bulk clutter into flexible packaging. Captures reaching into a mixed bin, grasping small rigid parts, and inserting them into limp plastic bags — the combined rigid-grasp plus deformable-bag handling that real packing lines demand.
About This Dataset
An egocentric off-the-shelf (OTS) video dataset is a set of first-person recordings that has already been collected and can be licensed for AI training. The footage is captured with wearable cameras, so it shows tasks the way the person doing them sees them: hands, tools, objects and environment in one continuous view.
Because the data already exists, you can open samples, check that the activities, camera setup and video quality match your model, and start training sooner than with a new collection project. If you need a scenario, location or device that the samples do not cover, our custom egocentric video data collection service records it for you. Other data types are available as OTS datasets for multi-camera video, images and audio.
Real first-person footage across household, fabrication, agriculture and industrial tasks
Samples you can preview before you request a dataset
Per-video specifications: volume, participants, resolution and frame rate, camera rig, camera, format and scalability
Annotation and customisation available on top of existing data
Use Cases
First-person video supports several kinds of AI models. These are the most common.
Egocentric video shows how people grasp, re-grasp and handle objects from the hands' point of view, which is the viewpoint a robot learns from when training manipulation policies.
First-person recordings of everyday and industrial tasks give action-recognition and activity-understanding models real, varied examples across household, fabrication, agriculture and industrial settings.
Wearable-camera footage ties what a person sees to what they do next, useful for embodied agents, task planning and models that connect video with language.
Step-by-step workflows such as welding, soldering, packing or cleaning help models learn how multi-step tasks unfold and where mistakes happen.
Need labels as well? See our egocentric video annotation services.
Answers to common questions about off-the-shelf egocentric video data.
An Egocentric Off-the-Shelf (OTS) dataset is a pre-collected first-person video dataset that is immediately available for AI model training, research, robotics, embodied AI, and computer vision applications without requiring a custom data collection project.
Egocentric OTS datasets are widely used in robotics, manufacturing, logistics, healthcare, retail, AR/VR, autonomous systems, industrial automation, and academic AI research.
Datasets may include household activities, warehouse operations, industrial workflows, retail interactions, healthcare procedures, cooking, cleaning, object handling, navigation, and other real-world human activities captured from a first-person perspective.
Yes. Existing datasets can often be extended with additional annotations, metadata, or supplemented with custom data collection to meet specific AI training requirements.
Yes. Video annotation services such as object tracking, bounding boxes, semantic segmentation, action labeling, event tagging, and quality validation can be provided alongside off-the-shelf datasets.
Browse the samples above and open any video to see its description and specifications. Then use the contact button on this page and tell us your use case, the activities you need and the volume you are looking for. We will reply with the matching dataset options.
More OTS Datasets
Need another data type? Browse our multi-camera, image and audio collections, or see all OTS datasets.
Get Started Today
We capture real-world first-person data for robotics, embodied AI, computer vision, and autonomous systems.
