Object Detection Using Convolutional Neural Network To Identify Popular Fashion Product

Abstract

Up to 95 million of photos are uploaded every day to Instagram as the biggest photos-sharing platform in the world. People share photos of their daily activities, hobbies, opinion and outfits in their Instagram post. This phenomenon opened a new visual discovery opportunity where unwritten information can be extracted using computer vision technology. One can extract the insight using Convolutional Neural Network (CNN) by applying object Detection process to recognise object in an image as a part of quantification process and knowledge extraction. In this paper, we re-train the CNN model using top-head fashion accessories dataset, represented by veil, eyeglasses and hat, to recognise the use of those item in each Instagram Image posts gathered as a dataset from 3 representative cities in Indonesia to identify the most popular accessories used in each representative city. The benefit of this model is the ability to identify popular top-head fashion accessories used in a particular area automatically, utilize it as basis information for company in fashion industry to better understand market, thus, increase the accuracy in decision making.