Real-Time Fashion Landmark Detection for Fashion Analysis and Understanding

The goal of this research is to develop an algorithm which allows Dataperformers to detect and match the apparel present in the video with the apparels in their inventory. To uniquely identify apparels it is very important to match them based on certain key areas. Fashion landmarks are the position of functional key-points defined on the fashion items such as the corners of a neckline, hemline, and cuffs. In this research we propose an algorithm which can find fashion landmarks in real time. We do so by learning to extract features which can provide location information as well as visibility of the fashion landmarks. This algorithm can be the core component to design the in-video store where a video consumer can purchase apparels from the videos. It is one of its kind approach for ecommerce.

Faculty Supervisor:

Jia Yuan Yu

Student:

Partner:

Dataperformers Company Inc.

Discipline:

Engineering

Sector:

Retail trade

University:

Concordia University

Program: