E-commerce is growing rapidly across the world. The major advantage that e-commerce websites have over retail stores is the abundance of user data, which can be used to make recommendations for the user. Retail stores need to gain a better understanding of their users.
As a solution to this problem faced by retailers, our team aimed to build a basic system that uses object detection to identify the clothing style worn by the customer. Clothing recognition can be used for smart customer analysis and for building their clothing profile. Once the clothing style has been identified, the system can suggest similar products from the store's database.
The solution implemented uses occasion wear to distinguish between various clothing styles. We use transfer learning to train an object detection model to identify clothing items in the input image and predict its style. This information is then passed onto a recommendation system that recommends similar clothing items on the basis of the predicted style.
In this series, I will go into the details of building this system. The complete detection and recommendation system was developed with @gauravaidasani , where I mostly focused on the style detection part, and hence it will be mostly focused on my contribution.