- Joined
- May 13, 2025
- Messages
- 1,631
- Online time
- 6d 13h
- Reputation
- 3,398
LooksWise - Facial Perception Research Platform (Beta)
A friend of mine is running a public beta for a facial perception research project. He would appreciate it if you guys tested it out.
How it works:
Upload a neutral, front-facing photograph. Rate other participants' photos, and in return your own photo is rated by other users.
As ratings accumulate, you unlock a report describing how your photo was perceived.
Isn't this just PhotoFeeler?
Similar idea, but the methodology is more controlled. It controls for lighting, facial expressions, and other occlusive variables wherever possible, including body, hairstyle, clothing, etc.
What if people don't rate carefully?
There's quite a few quality-control signals including minimum viewing times, consistency checks, calibration tasks, and rater reliability metrics. So the idea is to minimise the influence of rushed or inconsistent participants by weighing each rate differently.
What does the report show?
Current beta:
- Overall composite score
- Rating distribution
- Participant agreement
- Statistical confidence
- Preliminary results before completion
- Full report after sufficient ratings have been collected
What's the catch?
The platform is free because anonymous, consented rating data is used to develop the platform's machine-learning model. Experimental AI predictions and feature analysis will become available as the dataset grows.
First 50 beta users who contribute to the beta site will be given extra report analysis access on the full site including facial measurements and some other things. All cool stuff.
And as always, data is treated in accordance to privacy laws and only the minimum will be retained for the research data. All data will be irreversibly anonymised. Data deletion is provided upon request. Photos will only be used for the purpoaes that users explicitly consent to upon sign up.
Few examples of what a report page looks like:
A friend of mine is running a public beta for a facial perception research project. He would appreciate it if you guys tested it out.
How it works:
Upload a neutral, front-facing photograph. Rate other participants' photos, and in return your own photo is rated by other users.
As ratings accumulate, you unlock a report describing how your photo was perceived.
Isn't this just PhotoFeeler?
Similar idea, but the methodology is more controlled. It controls for lighting, facial expressions, and other occlusive variables wherever possible, including body, hairstyle, clothing, etc.
What if people don't rate carefully?
There's quite a few quality-control signals including minimum viewing times, consistency checks, calibration tasks, and rater reliability metrics. So the idea is to minimise the influence of rushed or inconsistent participants by weighing each rate differently.
What does the report show?
Current beta:
- Overall composite score
- Rating distribution
- Participant agreement
- Statistical confidence
- Preliminary results before completion
- Full report after sufficient ratings have been collected
What's the catch?
The platform is free because anonymous, consented rating data is used to develop the platform's machine-learning model. Experimental AI predictions and feature analysis will become available as the dataset grows.
First 50 beta users who contribute to the beta site will be given extra report analysis access on the full site including facial measurements and some other things. All cool stuff.
And as always, data is treated in accordance to privacy laws and only the minimum will be retained for the research data. All data will be irreversibly anonymised. Data deletion is provided upon request. Photos will only be used for the purpoaes that users explicitly consent to upon sign up.
Few examples of what a report page looks like:
