[offtotheraces]

I’ve had this business idea for years but I would be torn apart in todays world if I tried to launch it:

Build an ML model trained on actual images of women’s breasts and bras that are verified to fit them well. Once the core model is built a separate, a smaller size on-device model is built to be the model users actually interact with. A user submits an image of breasts and is told right away which bras in which sizes would be the best for them. Photos containing faces would be blocked from being uploaded. No photos would leave the device unless a user opted in to do so in order to help improve the model.

Monetization is straightforward: affiliate fees for directing users to buy bras at retailers, plus sponsored placements in search results at some point.

Biggest concern is how to come up with the initial data set, but I’m fairly certain there’s a solution there.

Am I crazy or is this as obvious as it seems to me? ThirdLove sucks, as do calculators like this.

[AussieWog93]

I'm not a woman, but do have over a decade of experience looking at pictures of breasts.

To be honest with you, I don't think it would be possible to properly parameterise a boob based on nothing but still images. They all have different centres of mass, different elasticities, different shapes and different textures. Not to mention that different women engage in different activities that require different levels of support.

You might be able to get a fit using 3D scanners plus some sort of standardised jiggle generator, but by that stage you might as well just have a skilled consultant hand you a few different pairs to try on.