ClosetAI

Homepage

Inspiration

ClosetAI draws inspiration from three key sources that aim to simplify your life!

Firstly, it takes a page from Apple's innovative Photos application, which uses facial recognition to categorize people in your photos. Just as Apple's technology makes it effortless to find pictures of loved ones, ClosetAI leverages similar image recognition techniques to identify and organize the clothing items you wear in your photos.

Secondly, the inspiration behind ClosetAI arises from the everyday challenge we all face: selecting outfits from our overflowing wardrobes. We understand the hassle of sifting through drawers, unfolding and refolding clothes, and the frustration of not finding that perfect ensemble. ClosetAI was born out of the desire to eliminate this hassle by turning your camera roll into a virtual wardrobe. Now, you can effortlessly search, mix, and match your clothes to create stylish outfits, simplifying your daily routine and ensuring you always look your best without the wardrobe chaos. It's all about using technology to make your life more convenient and your style choices more effortless.

Third, the idea for ClosetAI was also inspired by the fact that many times when we're thinking about creating an outfit, we have an idea of using something we've worn before. So, we find ourselves scrolling through our camera roll, searching for photos where we're wearing those clothes. This all-too-familiar scenario prompted us to create a solution that not only streamlines your wardrobe but also helps you effortlessly locate those beloved outfits from your past, bringing your fashion visions to life with ease.

What it does

Our product, ClosetAI is a mobile application designed to simplify and enhance your fashion experience. Here's how it works:

1. Camera Roll Integration: ClosetAI seamlessly integrates with your smartphone's camera roll or photo gallery, where you likely have numerous photos of yourself wearing different outfits.

2. Automatic Clothing Recognition: The app employs advanced image recognition technology to identify the clothing items you're wearing in your photos. It can distinguish between various clothing types like shirts, pants, dresses, and shoes.

3. Virtual Wardrobe Creation: After analyzing your photos, ClosetAI compiles the recognized clothing items into a virtual wardrobe within the app. Each item is categorized, making it easy to find later.

4. Outfit Customization: Using your virtual wardrobe, you can mix and match different pieces of clothing to create unique outfits. The app provides a user-friendly interface that allows you to experiment with different combinations, ensuring you always look your best.

5. Search and Planning: ClosetAI comes with powerful search and planning features. You can search for specific clothing items and filter outfits by style.

How we built it

ClosetAI is based on the following backend processes:

1. Images are imported from the camera roll or photo gallery of the user’s device.

2. Imported images are fed to a Machine Learning (ML) model which detects human outlines and provides bounding box coordinates for human’s silhouettes.

3. A cropped image of a singular person is extracted from the image and exported as a new image.

4. This new image is then fed to a face recognition ML software to determine whether the selected person is the subject of interest.

5. If the image does indeed contain the subject of interest, a clothing recognition model is run to determine the articles of clothing which are worn in the photo.

6. These clothing items are then saved in the virtual wardrobe whose functionality can be accessed and manipulated by the user through the front end GUI.

Challenges we ran into

There were various challenges we ran into while developing ClosetAI which proved difficult to overcome primarily as a result of the intense time restriction:

1. Project Breakdown: One of the hardest stages of the process was breaking down the functionality of ClosetAI into logical and practical sub-tasks. That is, we began with the complete application functionality and began working backwards to determine all the components which we would need to implement.

2. Frontend to Backend Data Transfer: We encountered a challenge when it came to transferring user-inputted photos from the front end to the back end of our system as a result of inexperience with this process.

4. Data collection: We encountered a significant challenge due to the sheer size of the data, which was initially inaccessible. We had to download, unzip, and upload the images, a process that consumed seven hours of valuable time.

5. Model Training: While attempting to train our model we ran into the unforeseen and unfortunate issue that something had gone wrong and the training was not able to be completed. Thankfully, we were able to locate the issue, which was related to the image formatting, and re-train the model successfully.

6. Holistic Implementation: After completing the development of each individual sub-task, combining them all into one application proved to be challenging and forced us to deal with multiple unexpected setbacks.

Accomplishments that we're proud of

We are incredibly proud of several key accomplishments that showcase our team’s grit and perseverance as well as define the success of ClosetAI:

1. Image Recognition Model: We achieved a high level of accuracy in recognizing and categorizing clothing items in user photos using our own model which we built, trained and tested. Our advanced image recognition technology allows users to build their virtual wardrobe effortlessly.

2. Human detection and Face Recognition: Finding two pre trained, open source, ML models which could be directly pipelined into our project, was an unexpected yet extremely helpful advancement for our product development. Locating these models took a significant amount of time due to the specific requirements which we had, especially for the face recognition program. However, this time spent definitely proved valuable and worthwhile.

3. User-Friendly Interface: Our team is proud of our easy-to-use GUI, allowing us to successfully showcase the functionality of our application.

What we learned

Full Stack Web App Development: The ClosetAI project provided valuable insights into the intricacies of building a full-stack web application. This encompassed various aspects, from front-end user interfaces to back-end server systems, databases, and ML models. The project taught us the fundamental skills of how to integrate these components to create a cohesive and user-friendly application. This newfound knowledge in full-stack development equips our team with the skills to tackle complex, end-to-end software projects effectively.

Time Optimization in Strict Timeframes: Working on ClosetAI within a very tight timeframe challenged our ability to optimize time and resources efficiently. We learned the importance of effective project management, task prioritization, and agile development methodologies. This experience allowed us to develop strategies for meeting project milestones and delivering a high-quality product under strict time constraints. These time optimization skills are valuable in any project and will be beneficial for future endeavors.

What's next for ClosetAI

Complete Implementation: We are working towards the completion of the integration of all parts of the system. Specifically, the Backend feeding data to the ML models, which work collaboratively amongst each other, and then returning the labeled and classified data to the Backend.

Style Inspiration: The app is intended to be able to also provide style inspiration based on your existing wardrobe. It might suggest outfit ideas or combinations you haven't tried yet, helping you make the most of your clothing collection.

Fashion Style Filtering: We are working towards running our clothing recognition software on various fashion and style trend datasets. By doing so we can extract the most popular articles of clothing or accessories related to each style and implement a "style" filter on our app.

Accessories: Our team is actively improving our clothing recognition model to include recognition of bracelets, necklaces, earrings and rings.