Doodle2Clothing : a clothing design recognition and searching model from a doodle drawing

Abstract

In recent years, pattern recognition has made significant progress; as computers can now identify and classify patterns of objects from various types of visual sources. These substantial forward strides in image and pattern recognition research can help users in ways that can improve the end-user experience for a lot of applications. In our research, we would like to solve a problem that shoppers usually face: finding a clothing based product with a specific design or pattern the shopper has in mind. With the use of Computer Vision and pattern recognition, our system will be able to recognize designs etched onto the product after classifying the type of clothing from doddle sketches provided by the users and we will be able to get the optimal images or products from the internet for the users. We will be training our model by collecting doodle sketches from end users by showing them an image of a clothing product and the model will extract features by comparing similarities between the sketch and the original picture based on its shape, color, design, etc and we expect to generate a search query which will be executed to find the product which is the closest resemblance to the doodle drawn by the user. With the help of our model, shoppers will be able to find their desired cloth without the hassle of excessive browsing; thus we will be enhancing Human-Computer interaction and provide a better and easier shopping experience for shoppers.