Curated datasets · Egocentric human activity

Building theworld’s largestcurated 1st-personPOV datasets.

First-person data for human life activities — curated from published sources, synthetic examples, and planned glasses-POV capture.

Explore the datasets

Meet HumanCapture

A planned $5 reward for each accepted first-person POV datapoint captured with glasses and annotated.

HumanHud.ai

Transcribe daily life into a calendar and transcript that AI agents can use.

HumanCapture

First-person concept

Loading first-person preview

See it through the wearer’s eyes.

A continuous first-person clip preserves the action and its context.

Viewpoint: First-person · Glasses POV
Media: Synthetic concept video

Explore everyday life

8 categories

Choose a category to explore five everyday activities.

Egocentric video · Sensor data · Human annotation

Scroll to explore

01Egocentric models02Audio + vision03Developer APIs04Paid contributions

01 / Datasets & models

Human activity. A shared intelligence library.

We’re building a library of AI models and datasets that categorize what people do. The same intelligence will power HumanHUD and be available to developers through APIs.

The dataset roadmap

Start with daily life: 10; Capture focused datasets for familiar activities.
Add depth and variety: 100; Expand tasks, environments, and ways of doing them.
Connect more of life: 1,000; Cover more routines, skills, and situations.
Build toward broad coverage: 10,000; Grow the library as quality and coverage are validated.

Dataset targets, not current catalog counts. Our long-term ambition is 99.9% coverage of human activity; the scope and measurement still need to be defined. Models and developer APIs are in development.

First collection focus / Everyday life

Eating
Sleeping
Cooking
Gym & exercise
Walking
Cleaning
Personal care
Working & studying
Shopping
Rest & recreation

Ten starting categories to shape the first datasets. Each collection will define the activity, requested data, and annotation requirements before it opens.

Find what exists. Build what’s missing.

Curate published data

Find useful datasets across the internet, preserve their provenance, and respect each publisher’s access and usage terms.

Fill gaps with synthetic data

Develop generated examples for activities that lack suitable data. Keep synthetic records identified and evaluate them against real observations.

Learn from lived experience

Grow collections with opt-in egocentric video and sensor data from HumanHUD and HumanCapture, paired with human annotations.

One library. Two ways to build.

HumanHUD brings activity intelligence to AR glasses. Our planned APIs will give developers access to the same model library for their own applications.

Browse existing dataset sources Discuss developer access

HumanCapture / In development

Your experience. Better training data.

HumanCapture is the app we’re building to pay people $5 for each accepted first-person POV datapoint captured using compatible glasses and fully annotated. Choose a dataset task, record the activity and requested sensor data, and label what happened.

Proposed base bounty

per accepted, fully annotated datapoint

Capture and annotation are two parts of one contribution. We’re exploring a split that rewards the people who complete each part.

Capture: $3; Record the requested activity and sensor data.
Annotate: $2; Label the activity and add the context the dataset needs.

Proposed split · $5 total when you complete both parts

01
Choose a task
See which datasets need contributions, what to record, and the bounty offered before you begin.
02
Capture a datapoint
Use compatible glasses with a supported capture app or HumanHud.ai integration to record the requested first-person video and sensor data.
03
Add the annotations
Annotate your own submission, or let another eligible contributor earn the annotation portion. You can also choose annotation-only work.
04
Submit for review
Accepted capture and annotation work earns the posted reward. Each task defines the evidence and quality requirements.

Collect what’s needed. Reward the work.

Tasks that require more effort may offer higher bounties. Once a dataset has enough accepted datapoints, its collection closes. Available work and reward amounts are specific to each task.

You choose what to contribute. Collection requires participant permission, and annotation access is limited to eligible contributors for that task.

HumanCapture, contributor payouts, and the $3 / $2 split are in development. Final task requirements and payment terms will be published before collection opens.

HumanHUD / A connected product

In development · A product by Matha Labs

HumanHud.ai.

Your life, in a calendar and transcript.

We’re building HumanHud.ai to transcribe a person’s daily life through compatible glasses into a calendar and transcript that AI agents can use. Our integration direction includes agents such as Meta Muse, Hermes, OpenClaw, and OpenAI Dot; these integrations are planned.

Visit HumanHUD.ai Inside the product

humanhud.aiIllustrative HUD

Where we’re taking it

Our first ten datasets focus on everyday activities from the first-person view, such as cooking, exercise, reading documents, and computer work. HumanHud.ai is intended to turn that activity context into a useful personal record. HumanCapture is our planned app for paid, opt-in glasses-POV capture and annotation; evaluated models and developer APIs connect the shared library to future applications.

Platform capabilities

Capture capabilities vary by device. The current Meta Ray-Ban Display web interface is display and D-pad only; it does not have camera or microphone access. Activity capture requires a compatible platform or companion camera, and those integrations are in development.

Capture the context

Use a compatible first-person view, sound, and activity context to understand the moments of daily life.

Build a personal record

Organize what happened into a calendar and transcript, so useful moments can be found again.

Give agents useful context

Build toward permissioned access to a person’s calendar and transcript so connected AI agents can help with relevant context.

03 / Supporting intelligence

Three tiers. One human purpose.

First-person activity datasets are the foundation of our work. Starting with ten everyday collections, we plan to develop and evaluate models for developer APIs and HumanHUD, where local visual and audio intelligence and Cloud AI can help a person with the task at hand.

Human

You set the direction.

Your goals, choices, and actions give the system its purpose. Ask for help, guide the task, and see what your AI is doing on your behalf.

Intent · Guidance · Feedback

Local edge AI

Intelligence where you are.

Egocentric models combine first-person visual and audio context to interpret activities and nearby objects. Local AI supports timely feedback through the glasses.

Audio · Vision · Activity context

Cloud AI

Working on your behalf.

Your AI helps plan the task, works across tools, and acts on your behalf. It directs and coordinates local edge AI, using feedback from your environment to adapt.

Planning · Delegation · Coordination

How they work together

HumanSets the goal
Cloud AIPlans and directs
Local edge AIPerceives and acts
HumanSees the result

↳ Human feedback shapes what happens next. People who opt in can also contribute video, sensor examples, and annotations through planned HumanHUD capture integrations and HumanCapture to improve the shared library.

Imagine help with the task at hand

You ask for help assembling something. Through compatible AR glasses, HumanHUD uses the first-person view, sound, and activity context to understand the step you are on. Local intelligence interprets the scene, Cloud AI helps plan the next step, and the glasses show useful guidance. You remain in control of whether any example is contributed to a dataset.

Product direction · Illustrative scenario

Explore the products

04 / Research directions

Start with the data. Explore what it makes possible.

Start with the human activity. Understand the first-person context, find a useful way to help, and carry that knowledge into models that act in the physical world.

Inside the lab

First-person activity datasets

Start with ten datasets about everyday activities such as eating, sleeping, cooking, and going to the gym. Combine documented sources, clearly identified synthetic examples, and planned opt-in capture.

Activity understanding

How can egocentric models combine visual and audio cues to understand what a person is doing, what matters in the scene, and when help would be useful?

Spatial interfaces

How can people express intent, guide local actions, and understand what Cloud AI is doing on their behalf through glasses and spatial interfaces?

Much later: physical AI & robotics

Preserve the desktop arm, mobile robot, and wearable backpack concepts for later research. Human demonstrations could eventually inform physical assistance, after the dataset foundations and separate hardware validation.

05 / Enterprise data & consulting

Data for your models. Support for your next step.

Explore our curated directory of external dataset sources, or discuss a custom collection for your target activities. We also offer focused AI consulting built around your goals.

Discuss your project

Discuss an AI challenge.

Start with a free 15-minute introductory call with Revanth. We’ll discuss your training data needs or AI challenge and how Matha Labs can help.

Book a call (661) 372-5742