Image recognition app. AI-based skin care diagnostic imaging.

[juresesko]

Do you have problems with your skin. Pimples, redness, acne... But you don't think it's worth it going to a doctor.
Imagine having an app in which you would upload pictures of your face and an AI would find imperfections on your skin and based on those imperfections suggest you different skin care routines and products.
I can imagine a lot of people would use this app on a daily basis. They could update their pictures and AI would compare them and giving them some feedback on what else they could do.
This app could bring a lot of value to a lot of people.
If you look at it from CENTS perspective you are in the green.

The next question is how do you make something like that I am not a programmer and I don't know a lot about AI.
If you maybe know what book to pick up or who to connect with please leave comment.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[openwater]

Do you have problems with your skin. Pimples, redness, acne... But you don't think it's worth it going to a doctor.
Imagine having an app in which you would upload pictures of your face and an AI would find imperfections on your skin and based on those imperfections suggest you different skin care routines and products.
I can imagine a lot of people would use this app on a daily basis. They could update their pictures and AI would compare them and giving them some feedback on what else they could do.
This app could bring a lot of value to a lot of people.
If you look at it from CENTS perspective you are in the green.

The next question is how do you make something like that I am not a programmer and I don't know a lot about AI.
If you maybe know what book to pick up or who to connect with please leave comment.

An coworker and I actually made an attempt at this space. The good news for you is that there is lots of competition, meaning it's a great need. The custom skincare space is currently exploding.

The work involved before development is likely thinking through your unique selling proposition vs. some of the competition out there. Here are a few that we encountered:

Skin Care Diagnostic Tool | Clinique | Official Site

Use the Clinique skin care diagnostic tool to determine the right solutions for your skin care needs. Allergy Tested. 100% Fragrance Free.

www.clinique.com

Treat Acne, Fine Lines and More | Curology Custom Skincare

Powerful, personalized skincare formulas prescribed by real dermatology providers. Start the skin quiz to try it for 30 days.

curology.com

https://www.provenskincare.com/

MDacne - Custom Acne Treatment

MDacne analyses acne prone skin and provides real time information about the acne type and severity together with a custom personal acne treatment plan.

www.mdacne.com

Your personal clear-skin regimen | Walgreens - Cloned

Walgreens. Trusted Since 1901.

www.walgreens.com

Editing to add one more thought - a mistake we made was to actually try creating the creams based on the images of users, which is expensive and very regulated. My two cents is for you to develop the system for identifying someone's need based on a picture and go the route of getting acquired by an existing skincare company (it sounds like this was your plan anyway).

 

[Kasimir]

Do you have problems with your skin. Pimples, redness, acne... But you don't think it's worth it going to a doctor.
Imagine having an app in which you would upload pictures of your face and an AI would find imperfections on your skin and based on those imperfections suggest you different skin care routines and products.
I can imagine a lot of people would use this app on a daily basis. They could update their pictures and AI would compare them and giving them some feedback on what else they could do.
This app could bring a lot of value to a lot of people.
If you look at it from CENTS perspective you are in the green.

The next question is how do you make something like that I am not a programmer and I don't know a lot about AI.
If you maybe know what book to pick up or who to connect with please leave comment.

Definitely a great idea. Don't know a lot about Al myself but I don't think it's that difficult. You may need to implement some questions like age and so but otherwise I think it's an amazing idea.
The market is perfect and the timing is also quite great.
Good luck!

 

[OverByte]

The next question is how do you make something like that I am not a programmer and I don't know a lot about AI.
If you maybe know what book to pick up or who to connect with please leave comment.

I do think this is a good idea and as @openwater shared some competition its obvious this could do well.

Based on your comment of having no programming knowledge, you should know that this would certainly not be a trivial app to build. Even if you could find a model you could use to identify the skin conditions (which would be an exercise into itself especially if you don't know what to look for). You'd then need to build a polished app on top of it. You definitely can't just pick up a book and expect to feel your way through it.

If you are serious about this I'd strongly suggest finding someone technical to partner with even if you are going to outsource the development. I suggest this so you don't hire someone who butchers the job and you loose a bunch of money. If you are going to try to build this yourself without programming knowledge.. don't. Build a simpler app instead to get your feet wet.

Don't know a lot about Al myself but I don't think it's that difficult.

Before making a judgement on effort it's important to understand what is actually involved in doing this:

The end goal is to have a trained model to recognize skin conditions based on user uploaded photos. There are several ways to go about this:

1 - by far the easiest - find a company that already built such a model and licence their API

2 - find a massive database of skin condition images labelled with metadata indicating the condition
2a - choose an image classification model (I would probably recommend using something that is available through Google ML but you could also find a model yourself)
2b - take a portion of the dataset as training data and run through the model until you get to a good accuracy of classification. Make decision whether it's better to have false negatives or false positives for the particular application. This will involve a bunch of parameter tweaking which assumes you understand at least at some level how the model functions.
2c - once you are happy with performance of 2b, run the model over the other portion of the dataset you have (with known outcome and not used for training) this will be your test data set. If it works then you have some confidence in your model, if it doesn't it you need to repeat 2b.
2d - note that even if 2c works you need to understand whether your model is too biased based on the training/test data and won't actually work for real world results. You'll need to understand how to make an informed judgement based on the training/test data. If you have a bunch of clear cut images then it may give great results for the training data but be useless in real life
2e - test it with real results from user uploaded photos from people who already have a confirmed diagnosis. See if model makes the right one.

If neither 1 or 2 are available then wait for some dermatology company to make 1 or 2 available.

After you have a good model you can figure out how to build a highly polished app on top of it.

That is just how you would technically go about this, before this you'd have to figure out the business model, etc.

Note - I do not post this to discourage anyone but rather would like to share some information on how you would practically go about this. It seems like with AI/ML being a giant buzz word many aspiring software entrepreneurs assume you can just add AI to some problem and have it work. Any AI application starts with data and then a trained model. So from a technical standpoint you need to figure out how to get both of those before you can considering adding any sort of "intelligence" to your software. Most of the work associated with ML is finding and labelling data (massive amounts of data).

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[openwater]

Having some fun with this now...

If the time and expense of building an AI app (or hiring someone to do it - ML engineers are not cheap) is too high, you could get creative with delivering the value to your users.

For example, you could enlist a dermatologist to create a very simple "how to" guide on spotting skin characteristics via a selfie (red skin, shiny skin, dry skin, acne).

Now you can hire out the custom labelling of selfies to Mechanical Turk for cents per label, with them referring to the "how to" guide.

How are you planning to generate revenue from this? One thought is that you're left with a list of users, their emails, and their skin issues. Could be an interesting lead generation play for skincare companies.

This would be a whole lot cheaper than training an AI model, and the USP is that it's dermatologist developed (instead of a blackbox algorithm).

From experience, I can tell you that the amount of garbage pictures your model would have to deal with would be unbelievable. Blurry, dark, upside down, of a cat, of a wall, etc. The human element can pretty quickly provide feedback to say "please provide another picture".

 

[BellaPippin]

Not too long ago I remember a tool (not sure if an app or filter or what) that rated your skin based on a selfie. It was pretty thorough in the sense that sort of listed the things it found or lack thereof like "pimples: 0" age marks, skin spots, etc. I forgot the name but it was made to recommend products.

There's also businesses like Prose (for hair) that formulate a product based on a quiz, and actually one for skin based on a quiz too. Then another one where supposedly a dermatologist looks at a few selfies and recommends a treatment. Sorry I'm all over the place, my memory kept naming them lol. I see them most on IG ads.

 

[Kasimir]

Not too long ago I remember a tool (not sure if an app or filter or what) that rated your skin based on a selfie. It was pretty thorough in the sense that sort of listed the things it found or lack thereof like "pimples: 0" age marks, skin spots, etc. I forgot the name but it was made to recommend products.

There's also businesses like Prose (for hair) that formulate a product based on a quiz, and actually one for skin based on a quiz too. Then another one where supposedly a dermatologist looks at a few selfies and recommends a treatment. Sorry I'm all over the place, my memory kept naming them lol. I see them most on IG ads.

Yeah, I think you mean something like this: This App Can Track How Well Your Skin-Care Products Are Working

If they can rate your skin and tell you if your current products are working or not. Then it's not that hard to transfer that into your own skincare products. But as said before I'm not a programmer.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[Jon L]

This is a very difficult programming problem to solve. You will need a number (5-10?) of HIGHLY skilled programmers that will want $250k/year or so in salary, each. Plus dermatology consultants. Plus a lot of hardware. It will probably take several years of work to build it, and there is a decent chance that what you build won't work well enough. So, millions of dollars. And there are big-name companies that are trying to do the same thing.

Unless you have some secret sauce you haven't mentioned (meaning you personally have an idea for how the software would work technically that no one else has thought of), I wouldn't do it.

 

[Kasimir]

This is a very difficult programming problem to solve. You will need a number (5-10?) of HIGHLY skilled programmers that will want $250k/year or so in salary, each. Plus dermatology consultants. Plus a lot of hardware. It will probably take several years of work to build it, and there is a decent chance what you build won't work well enough. So, millions of dollars. And there are big-name companies that are trying to do the same thing.

If you build it from scratch it could take years. But if you just talk to some companies which already made it but aren't using it in your way. You could team up and give them either royalty on every product or a part in your company. The only problem would be is that they copied your idea and do it without you.

 

[juresesko]

An coworker and I actually made an attempt at this space. The good news for you is that there is lots of competition, meaning it's a great need. The custom skincare space is currently exploding.

The work involved before development is likely thinking through your unique selling proposition vs. some of the competition out there. Here are a few that we encountered:

Skin Care Diagnostic Tool | Clinique | Official Site

Use the Clinique skin care diagnostic tool to determine the right solutions for your skin care needs. Allergy Tested. 100% Fragrance Free.

www.clinique.com

Treat Acne, Fine Lines and More | Curology Custom Skincare

Powerful, personalized skincare formulas prescribed by real dermatology providers. Start the skin quiz to try it for 30 days.

curology.com

https://www.provenskincare.com/

MDacne - Custom Acne Treatment

MDacne analyses acne prone skin and provides real time information about the acne type and severity together with a custom personal acne treatment plan.

www.mdacne.com

Your personal clear-skin regimen | Walgreens - Cloned

Walgreens. Trusted Since 1901.

www.walgreens.com

Editing to add one more thought - a mistake we made was to actually try creating the creams based on the images of users, which is expensive and very regulated. My two cents is for you to develop the system for identifying someone's need based on a picture and go the route of getting acquired by an existing skincare company (it sounds like this was your plan anyway).

Definitely a great idea. Don't know a lot about Al myself but I don't think it's that difficult. You may need to implement some questions like age and so but otherwise I think it's an amazing idea.
The market is perfect and the timing is also quite great.
Good luck!

Thanks for the suggestions. I don't know a lot about it and the links you posted will provide some more information about what I want to start.
The program would in some kind of way recommend which skin care cream by an outside company would work the best. My app would a get a cut from the sales price, companies would pay us to be recommended and maybe there could a Subscription based service as well.

I do think this is a good idea and as @openwater shared some competition its obvious this could do well.

Based on your comment of having no programming knowledge, you should know that this would certainly not be a trivial app to build. Even if you could find a model you could use to identify the skin conditions (which would be an exercise into itself especially if you don't know what to look for). You'd then need to build a polished app on top of it. You definitely can't just pick up a book and expect to feel your way through it.

If you are serious about this I'd strongly suggest finding someone technical to partner with even if you are going to outsource the development. I suggest this so you don't hire someone who butchers the job and you loose a bunch of money. If you are going to try to build this yourself without programming knowledge.. don't. Build a simpler app instead to get your feet wet.



Before making a judgement on effort it's important to understand what is actually involved in doing this:

The end goal is to have a trained model to recognize skin conditions based on user uploaded photos. There are several ways to go about this:

1 - by far the easiest - find a company that already built such a model and licence their API

2 - find a massive database of skin condition images labelled with metadata indicating the condition
2a - choose an image classification model (I would probably recommend using something that is available through Google ML but you could also find a model yourself)
2b - take a portion of the dataset as training data and run through the model until you get to a good accuracy of classification. Make decision whether it's better to have false negatives or false positives for the particular application. This will involve a bunch of parameter tweaking which assumes you understand at least at some level how the model functions.
2c - once you are happy with performance of 2b, run the model over the other portion of the dataset you have (with known outcome and not used for training) this will be your test data set. If it works then you have some confidence in your model, if it doesn't it you need to repeat 2b.
2d - note that even if 2c works you need to understand whether your model is too biased based on the training/test data and won't actually work for real world results. You'll need to understand how to make an informed judgement based on the training/test data. If you have a bunch of clear cut images then it may give great results for the training data but be useless in real life
2e - test it with real results from user uploaded photos from people who already have a confirmed diagnosis. See if model makes the right one.

If neither 1 or 2 are available then wait for some dermatology company to make 1 or 2 available.

After you have a good model you can figure out how to build a highly polished app on top of it.

That is just how you would technically go about this, before this you'd have to figure out the business model, etc.

Note - I do not post this to discourage anyone but rather would like to share some information on how you would practically go about this. It seems like with AI/ML being a giant buzz word many aspiring software entrepreneurs assume you can just add AI to some problem and have it work. Any AI application starts with data and then a trained model. So from a technical standpoint you need to figure out how to get both of those before you can considering adding any sort of "intelligence" to your software. Most of the work associated with ML is finding and labelling data (massive amounts of data).

I have never though about option 1. Before you guys answered I didn't think something like that already existed, honestly I think there is still space for something like that. Maybe a little modified version that wouldn't rely on AI 100% at the start.

Having some fun with this now...

If the time and expense of building an AI app (or hiring someone to do it - ML engineers are not cheap) is too high, you could get creative with delivering the value to your users.

For example, you could enlist a dermatologist to create a very simple "how to" guide on spotting skin characteristics via a selfie (red skin, shiny skin, dry skin, acne).

Now you can hire out the custom labelling of selfies to Mechanical Turk for cents per label, with them referring to the "how to" guide.

How are you planning to generate revenue from this? One thought is that you're left with a list of users, their emails, and their skin issues. Could be an interesting lead generation play for skincare companies.

This would be a whole lot cheaper than training an AI model, and the USP is that it's dermatologist developed (instead of a blackbox algorithm).

From experience, I can tell you that the amount of garbage pictures your model would have to deal with would be unbelievable. Blurry, dark, upside down, of a cat, of a wall, etc. The human element can pretty quickly provide feedback to say "please provide another picture".

Not too long ago I remember a tool (not sure if an app or filter or what) that rated your skin based on a selfie. It was pretty thorough in the sense that sort of listed the things it found or lack thereof like "pimples: 0" age marks, skin spots, etc. I forgot the name but it was made to recommend products.

There's also businesses like Prose (for hair) that formulate a product based on a quiz, and actually one for skin based on a quiz too. Then another one where supposedly a dermatologist looks at a few selfies and recommends a treatment. Sorry I'm all over the place, my memory kept naming them lol. I see them most on IG ads.

This is one of the obstacles I would be facing. I don't know how to deal with low quality pictures. Maybe there could be comparison photos on the app and let the user decide if this is the right skin condition.
Revenues would come from different ways from sale %, add space, product recommendations and maybe some subscription service.

This is a very difficult programming problem to solve. You will need a number (5-10?) of HIGHLY skilled programmers that will want $250k/year or so in salary, each. Plus dermatology consultants. Plus a lot of hardware. It will probably take several years of work to build it, and there is a decent chance that what you build won't work well enough. So, millions of dollars. And there are big-name companies that are trying to do the same thing.

Unless you have some secret sauce you haven't mentioned (meaning you personally have an idea for how the software would work technically that no one else has thought of), I wouldn't do it.

Do you know it takes so much at the raw start.
Do you think I could pre-sell the app to investors to get the starting capital. Start with marketing right away and make the demo without the AI and only photos to compare.
Tech is getting better and better I think that you could do it for less money.

If you build it from scratch it could take years. But if you just talk to some companies which already made it but aren't using it in your way. You could team up and give them either royalty on every product or a part in your company. The only problem would be is that they copied your idea and do it without you.

Great thinking!

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[BellaPippin]

I don't know how to deal with low quality pictures.

Bet your bum most of the people will NOT send a good lit picture. You gotta have something like when you deposit a check and that kinda stuff that says "move closer" "too dark" etc. Or that one that has the little shape of the face and you align your face with it.

 

[Gallium]

Sounds kind of like MD Acne