By :- Danial Khan, Abhi Sharma, Suleiman Najim, Payas Hasteer
Problem: Ecommerce is one of the biggest industries in the world today and online apparel shopping has been at the forefront of this exponential rise of online ecommerce. Although ecommerce has made people’s lifes easier and more convenient, the process can still be enhanced particular when it comes to searching for products.
Alot of times people scroll through social media and see their friends or a celebrity wearing the perfect outfit. The problem is that the brand or product name can be unknown and therefore finding it online can be daunting. To combat this, we created a machine learning model that takes an image of an apparel as the input, predicts the category of the clothing/apparel and recommends the same or similar items that are in the same category. We believe that machine learning was the best and only way to accomplish this goal, as this is an image classification problem and we cannot explicitly the program to predict a clothing piece in an image as there are too many edge cases, so the best way is to use ML and left our model learn from our data so that it can predict images that have never seen before.
Overview: The Model predicts the apparel category the user-inputted picture belongs to. These apparel categories include: Men's shorts, Men's T-shirts, Men's Formal Suits, Women Dresses, Heels, Sunglasses etc.
Approach:
- Train: Model infuses 60% of all web-scraped images from Myers (online fashion platform) to train.
- Validate: Using 20% of the remaining images the model was validated.
- Test And with the remaining 20% the best model achieved an accuracy of 94.067%.
Future Scope: The team aims to train the model to brand differentiate and even follow fashion trends by providing priority to trending apparel clothing.
