Preparing a brazed jointMetalwork
Preparing a brazed joint
Positioning a torch and metal component at a fabrication bench.
Original AI-generated collection illustrationEGOCENTRIC DATA COLLECTION & ANNOTATION
Egocentric Collection.
The human perspective.
Egocentric data for the next generation of physical AI. Bring first-person video, human actions and hand–object interactions into one purposeful dataset.
Multicolor hand landmarks, camera poses, top-view path and scene points in one view.
Model-assisted preview - 142/160 camera frames registered - Relative scaleSource by Kaggle datasetDownload captioned video
CAPTURE THE HUMAN PERSPECTIVE
From the head camera
to the training dataset.
Connect purposeful first-person collection with hand, object and action annotation.
Plan your captureANNOTATION REVIEW EXAMPLE
0.0s / 40sAction captions. Hand keypoints. One timeline.
Review the caption and landmark overlay at the same moment. Select a frame or action segment to seek the preview below. Clip 00:00 starts at source 01:00.
VIDEO FRAME PREVIEWClip time 0.0s / Source time 60.0s
Action descriptionHand landmark overlayCamera pose / relative scale
Source by Kaggle datasetDESIGNED AROUND YOUR RESEARCHPhysical AIRoboticsAction understandingHuman–object interaction
M coding / FOOTAGE LIBRARY
Work in motion.
Detail in every frame.
Explore hand interactions across real tasks. Play each study to inspect the movement and its automatic labels.
Assembly hand tracking
Multicolor palm and finger overlays follow the hands through the assembly task. Automatic estimates require review.
Source by Kaggle datasetProcessed hand-overlay preview / 30 secDownload original videoSurface-work hand tracking
Repeated hand movement across a work surface, with estimated wrist and finger landmarks. Motion blur causes tracking gaps.
Source by Kaggle datasetProcessed hand-overlay preview / 20 secDownload videoM coding / 3D HAND MOTION
See the movement.
Understand the hand.
Follow the assembly task through an estimated 3D hand-pose view. The original viewpoint appears alongside, connecting each movement to the work.
Model-estimated hand motion. Hands are centered independently; this is not a calibrated scene reconstruction.
OUR COLLECTION CAPABILITIES
Local perspectives.
A connected network.
Explore our collection capability network across India, Bangladesh, Pakistan, the Philippines and Ukraine.
EGOCENTRIC COLLECTION NETWORK05 REGIONS
Collection capabilitiesMap data: Natural Earth
SELECTED REGION
India
Plan first-person task capture, hand–object annotation and quality review around your project brief.
Discuss collection in IndiaM coding / COLLECTION SCENARIOS
Different industries.
A human perspective.
Explore original synthetic scenes across everyday work. Compare collection views with hand and product segmentation, then inspect wrist and finger keypoints.
Preparing a brazed joint
Laying fabricTextiles & garments
Laying fabric
Material preparation before a sewing task.
Original AI-generated collection illustration
Applying a wood coatingWoodwork & furniture
Applying a wood coating
Following a brush and the hand applying a finish to a wooden panel.
Original AI-generated collection illustration
Folding a garmentGarment finishing
Folding a garment
Bimanual handling during garment folding and finishing.
Original AI-generated collection illustration
Operating a test stationElectronics
Operating a test station
Handling an electric racket at a small electronics testing station.
Original AI-generated collection illustration
Collecting extruded tubingPlastics & rubber
Collecting extruded tubing
Guiding flexible tubing from the production output into a coil.
Original AI-generated collection illustration
Positioning a printing framePaper & printing
Positioning a printing frame
Handling a printing frame and material at the press feeding table.
Original AI-generated collection illustration
Arranging ingredientsCommercial kitchens
Arranging ingredients
Preparation tasks in a professional kitchen environment.
Original AI-generated collection illustrationThese are original AI-generated illustrations, not customer recordings. Segmentation uses SAM 2.1 with hand and product prompts; keypoints are automatic estimates. These are review examples, not validated training labels. Glove contours are shown where reliable joint estimates are unavailable.
02 / DATA ANNOTATION
From first-person view
to structured labels.
Start with the human view. Define the actions, objects and moments your model needs to understand.
MODEL-ASSISTED FRAMEThe same frame.
The same frame.
A clearer view of motion.
Compare the original frame with multicolor hand landmarks from the video model. Each finger chain has its own color. Overlapping hands and hidden joints still need human review.
- Define the task and capture viewpoint
- Label hands, objects and interaction events
- Review visibility, occlusion and action timing
03 / CLEAR PROJECT PRICING
Two services.
One connected workflow.
Commission new footage, annotate your existing videos, or combine both. Each service is scoped separately.
Plan your collection budget.
Choose the camera, quality and volume for your brief. Adjust your planning rate and requested volume discount.
Mono$7starting / agreed unit
iPhone$10starting / agreed unit
GoPro$15starting / agreed unit
110,000
Starting rates vary by camera, client, task and product.Planning tiers suggest 10% from 1,000 units and 20% from 5,000 units. Edit this scenario; all discounts require agreement.Indicative planning total$700.00Effective planning rate $7.00 / unit · Requested savings $0.00
A unit is defined in your quote (for example, an accepted recording minute or a clip). This calculator is a planning scenario, not a confirmed tariff. Quality changes require a revised quote. Equipment provided by client / arranged by client. Annotation is priced separately.
Discuss collectionAnnotation pricing.
$0.015 / object
Published USD keypoint annotation reference. Video frame density, hand landmarks, action labels and specialist review are quoted separately.
- Hand keypoints and object regions
- Action and interaction labels
- Visibility checks and quality review
The reference rate is not a per-joint, per-frame or full egocentric dataset price. See published pricing
A FEW USEFUL ANSWERS
Before we start.
Can we combine collection and annotation?
Yes. Scope them together in your pilot brief, including capture tasks, participant consent, annotation guidelines, review criteria and delivery format.
Can we use our existing videos?
Share your footage requirements and an approved sample through the project discussion. We can scope annotation around the available viewpoints, image quality and usage rights.
What does the annotation sample show?
A real cleaning frame with automatic hand landmark estimates, alongside timecoded action descriptions in the featured video. Model outputs require human review, especially during occlusion.
How is the final price confirmed?
The published keypoint rate is a general reference. Collection figures are indicative and vary by project, product and client. The camera setup, billing unit and final rate are agreed in the quote. Equipment is provided or arranged by the client. Annotation, frame density, 3D work and quality review are scoped separately.