Can You Sell or License Egocentric Video to AI Labs? What It Is and How It Works
AITrainingDataBroker.com
Sell Egocentric
Video
to AI Labs?
to AI Labs?
What it is, who wants it, and who has to agree before it can be licensed
Jimmy Lin
Sell Egocentric Video to AI Labs?
A camera on a worker's head films a job exactly as that worker sees it: the hands, the tools and the order of the steps. That is egocentric video, and AI labs that build robots and smart glasses are looking for it. So can a company sell or license footage like that? We begin with 2 cameras.
The video
Full script
A camera on a worker's head films a job exactly as that worker sees it: the hands, the tools and the order of the steps. That is egocentric video, and AI labs that build robots and smart glasses are looking for it. So can a company sell or license footage like that? We begin with 2 cameras.
Most owners think video of their staff is too risky to license, and sometimes that is true. Still, look at the difference. The one on the wall sees the whole room, including every face, while the one on the head sees mostly the work, and that is what an AI lab wants to learn from. The catch is that the hands belong to a person, and so do the faces that walk into the frame. This is general information and not legal advice. First, let me tell you who I am.
I'm Jimmy. I have worked in data at Fortune 500 companies, as a data analyst, a data engineer and a software engineer, and I am trained in data science. I am also a Canadian military veteran, a commissioned officer specialized in data analysis and reporting.
If you want to sell your data, I am your broker and technical advisor. I bring your data to several AI companies, and I guide your team through the technical side, including compliance. AI companies pay my fee.
Already shopping your data around? Contact us anyway. We can still work through the terms with you and help you negotiate a better price.
If you are buying data for an AI lab, I find the right data and see it through to delivery, so you can focus on AI research.
3 kinds of AI learn from this footage. Robot teams use it to show a machine how a person's hands move through a task. Smart glasses must understand what the wearer is looking at, and the Ego4D research project even tests whether an assistant can answer where did I leave my keys. Screen recordings do the same for desk work, because they show every click. According to EgoVerse, a robot learning research paper, more human footage helps, but only when it matches what the robot is learning. By the way, we built a free data value calculator. It gives you an estimate, not a final offer, and the link is in the description.
To see how big this field already is, look at one research project. According to the Ego4D paper, researchers from a large tech company and 13 universities collected 3,670 hours of first-person video from 931 camera wearers in 9 countries. The paper says every wearer had to give informed consent, which is the first rule of the whole field. So the demand is real, and the question for you is what a company like yours could actually hold.
Companies hold more of this than they think. A field service company may have chest-camera clips of technicians at work. A training team may have a walk-through that an expert recorded and narrated on purpose, which is usually the easiest kind to start with because the person agreed. Office teams have screen recordings of orders and claims, and quality teams keep review recordings with the reviewer's note about what was done well.
It is not about the hours, and it is not about how sharp the picture is. It is about whether each clip comes with a note: the steps, the tool and the result. Remember the EgoVerse finding that more footage helps only when it matches what the model is learning. A repair clip that comes with its work order and the supervisor's score shows the lab what good looks like, while the same clip with no note is only a picture. That is why a sample needs its notes. But there is a bigger question to settle first: who is in the frame?
Every box on this screen is a person or a place that did not sign anything. The wearer can agree, but the customer in the kitchen has not agreed to anything, and neither has the person whose voice is on the audio. Some places treat a scan of a face as biometric data. In Illinois, the state's biometric privacy law says a company needs a written release before it collects one. In Europe, guidelines from the European Data Protection Board treat video of people you can identify as personal data. And in some US states, everyone in a conversation must agree before audio is recorded. The rules change by place, so your counsel reads the ones where each clip was made.
Some footage is simply left out, and we do not argue about it. A clip with no signed consent from the wearer is out. So is footage recorded in secret, or without a notice that the law required. Clips where a customer's face can not be blurred without ruining the clip are out, and so are recordings from clinics or customer sites that have their own privacy terms. That footage stays private and is never licensed to an AI lab. This will not fit every company, and I would rather say so now: footage kept only for safety or disputes was made for a narrower purpose, so counsel looks at it first. Many owners start with footage they made on purpose for training.
So what does cleaning mean? We de-identify the footage, which means we blur the faces, mute the voices of anyone but the wearer, and cover name tags, screens, house numbers and mail. It is slow work. The Ego4D paper reports that one team needed 780 hours of labor to clean 500 hours of video, and that a team at another university found personal details in 36% of its videos. Software finds most of it, and people check the result. That is why a company starts with a small sample, not the whole archive.
Now check your own footage with 3 questions, like a film crew checking a slate. Take them one at a time. Who wore the camera, and did that person sign a form that covers this use? Who else walks into the frame, and can they be blurred? And why was it recorded: for training, or only for safety and disputes? If you find a signed wearer, faces you can blur and footage made on purpose, you have a good first sample. Jot the answers down for 5 clips, and you will know where you stand.
Video files are large, so we begin with a small sample: a few clips from one wearer who signed. As your technical advisor, I guide your team through the export, the file format, the list that links each clip to its job note, the de-identification and compliance. As your broker, I show the sample to several AI labs, bring back their offers and work through the terms with you, and you make the final call. We move forward only after your team signs off.
If your team already films its work with consent, that footage may deserve a closer look.
If you run a company with 30 or more people, the data you already have could become a new revenue stream.
Want to find out?
Give our data value calculator a try at AITrainingDataBroker.com.
To keep going, watch Why No One Can Price Your Company's Data Without a Sample next. It is linked in the description.
If you have further questions, leave them in the comments, and I will answer them in the next video.
Thank you for watching.
I will see you next time.
More slides
AITrainingDataBroker.com
Can a Chip
Company License
Its Designs?
Its Designs?
What a design folder holds, who owns it, and which files cannot be licensed
Jimmy Lin
AITrainingDataBroker.com
Can a Pharma
Company License
Data to AI Labs?
Data to AI Labs?
What can be licensed, what stays private, and who has to approve it
Jimmy Lin
AITrainingDataBroker.com
Fintech and
Health Tech
Data for AI Labs
Data for AI Labs
What can be licensed, what stays locked away, and the rules to check first
Jimmy Lin
