Style Advice App Development with Features Based on Machine Learning Technology

It’s difficult to keep up with urban culture and the modern fashion trends that come with it. A paradox of modern shopping is that we often do it by ourselves, but we still want other people’s opinions and advice. That is why people rely on fashion advice apps. These apps combine the convenience of technology with human expertise to deliver opinions of friends and professional stylists to your phone or tablet.

Style advice apps come in different types: social fashion games where users create and vote on outfits,  tinder-like apps where users can like or dismiss an item by swiping the screen, fashion apps where users create and share moodboards (collages of images aimed at fashion inspiration), or a combination of all the above.

See our implementation of  Tinder-like cards for iOS on Github and Dribbble.

shopping app design

[Voting interactions for a fashion advice app designed at Yalantis. Check it out on Dribbble]

A lot of people get confused and do not see how style apps are different from shopping apps. This confusion is quite understandable because these two types of apps do overlap sometimes. For example, the Mallzee shopping app lets you put together an outfit, get your friends to vote on it, and purchase the clothes.

However, a principal difference between style advice apps and shopping apps is that style advice apps do not concentrate on selling items. They target people who want to reorganise their closets and get advice from friends or the fashion community. They might buy the items advertised in the app, but they don’t browse with the intent to purchase.

You might want to check out our article about 3 types of fashion shopping marketplaces.

Let’s imagine you decided to develop a fashion advice app. What features should it have? What is the most essential feature?  How can we make it feel simple and user-friendly? And, finally, how much would it cost to make an app of such a nature?

Features for good style apps

1. Personalized recommendations

Users can get style advice recommendations from a community of friends or professionals stylists. But how about style services and useful online platforms?

WeStyle is a good example of an app whose sole purpose is to give users a chance to hear a second opinion before making a purchase or selecting a look.

How to develop an app like WeStyle?

The users can upload a picture of their look and ask other members of the community to vote for or against it. They can also upload two pictures side by side and ask the community to vote for their favorite. This feature is considered to be among the most valuable according to the app reviews on the Internet.

WeStyle also has a cropping tool that helps users work on their pictures immediately before posting them, which is useful in case there are any changes to be made to the photo before it is posted.

WeStyle and similar apps wouldn’t be as helpful without a personalised recommendation system that lets the app show users the items they are likely to purchase based on their sense of style and lifestyle. This system is based on machine learning technology.

How does machine learning work?

Personalized recommendation systems are usually based on several different datasets that allow an app to divide items into different categories to make recommendations more relevant. For example, the popular fashion community Polyvore, which is available on both iOS and Android, have recently updated their app for iOS and now it is using the following types of division in their recommendation system, which you can try out on your iPhone or iPad. 

  • Content-based recommendations a selection of items based on what a particular user already liked.
  • Collaborative filtering – a selection of items based on what people with a similar taste already chose.
  • Complementary products– a selection of items based on what matches items that a user already chose.

Similar systems of filters are used in different e-commerce apps and they might be equally beneficial for the style advice mobile app because they help people focus on the most relevant items on the market.

These datasets help a style advice app narrow down preliminary results and after that the machine learning kicks in. Every time a user picks out an item or dismisses it, the app learns more about their style and preferences. This means that if a user is consulting  an app long enough, they will  get more accurate, personally tailored style recommendations.

2. The possibility to upload images and find a match on the Internet.

This is probably the second most important feature for style advice apps. When it comes to fashion or shopping on the Internet we are talking about hundreds of thousands of similar items. For any image-matching technology to work, algorithms should be able to accurately compare new and unknown images to older, identified images.

Creating a library of images can be done in partnership with a particular brand or a chain of stores, or you can sync the app with a database of images already available online. Then, you can implement image recognition functionality powered by machine learning techniques.

How does image recognition work?

Current machine learning techniques are far from perfect, even though the most sophisticated use neural networks.

Joint research run by Microsoft and Facebook (this research involved involving 300,000 images with containing 2.5 million different objects in them) showed that when a computer was competing with a team of people the machine’s  results only coincided with theirs only in 23% of the time. cases. Why? is that?

Machines (or, rather, special software) can easily detect colors or color combinations and basic shapes, but uneven backgrounds or odd angles prevents software from recognising objects in images.

In practical terms this means that if you were to develop an app that uses image recognition it would take a lot of time and money to make it worthwhile: you would have to compile a database of several million images (at least) and run it through the app or give out phones with a test-version of the app to people (probably, as many as 1 million!) so that they could use it over a relatively long period of time to take pictures, upload them to the app, and describe objects in these pictures. 

Image recognition algorithms need huge samples of data to improve performance.

How to develop a shopping app like Polyvore?

[The Polyvore app]

What is the alternative?  

Third-party image recognition APIs seem to be the best solution for image recognition. Developers can incorporate them into both mobile and web applications.

Among image recognition APIs currently available on the market there are a few that seem particularly fit for the task.

Cloudsight is a visual search and image recognition API that powers the Camfind app. Users take a picture or upload it to the library and Cloudsight returns the information that is already interpreted and provides a description of objects that are found in the image. The API is designed to be functional and accessible and serves as a back-end solution for image recognition.

Vufind Recognise is another cool image recognition API service which can even recognize brands from an image. They offer free plans and brand recognition APIs.

LTU Technologies  is one more high-quality image recognition service. It is set apart by several features including color searching and content tracking.

LTU offers different monthly plans (including free trials) and your choice of a plan will be defined by a number of factors.

Read also: How to develop a personal shopper app powered by data intelligence

3. Multiple filters and combinations

Popular fashion apps like Grabble and Mallzee are often referred to as “Tinder for fashion” and are all about an immediate “like or dislike” emotional response. Both apps gained popularity by mimicking a famous Tinder feature – users make a choice by simply swiping the screen left or right (in Grabble terminology users can “grab” or “throw” items). Users can share their “grabs” with friends via all major social media networks, messaging services, or a cloud service.

To be successful and up-to-date this type of app requires detailed filters that can divide all search results into groups according to types of clothes, color schemes, style or purpose (office clothes, casual clothes), brands, and, surely, price range - one of the most important filters.

These filters can work in two different ways. First, users can filter their search results as soon as they open an app by choosing a number of categories such as a type of item, size, color, style, and fabric (for example, users can pick “coat”, “wool’, “red”). But it is also possible for an app to remember what users liked most often and to offer a personalized selection of goods.

Style advice app concept designed at Yalantis

[Fashion advice app concept designed at Yalantis]

4. Push notifications

Another feature that users love is the ability to get push notifications if an item drops in price. If people see how an app can save them money, even if online shopping is not the app’s major feature, they are more likely to use it.

5. Social sharing

People want to be able to share their looks with a community and get feedback before they make a purchase at the store. Providing more built-in capabilities for social network interactions in e-commerce and style advice apps should increase an app’s popularity - it has been statistically proven that people are more likely to spend money on things that have been approved by their friends or professional style experts.

Image recognition and personalized recommendation systems are trending in the e-commerce and fashion industries. Yalantis can help you develop a style advice app to keep up with the times!

Read also: How much does it cost to develop an e-marketplace app?
4.1/ 5.0
Article rating
Remember those Facebook reactions? Well, we aren't Facebook but we love reactions too. They can give us valuable insights on how to improve what we're doing. Would you tell us how you feel about this article?
Check out Autoportal case study

Learn how to port online marketplace for cars to iOS and Android

Contact us

We use cookies to personalize our service and to improve your experience on the website and its subdomains. We also use this information for analytics.