Relative Attributes

Team

• Kanav Gupta
• Parth Partani
• Anurag Mehta

Introduction

Project involves obtaining a ranking function which can relate two or more images on strength of particular attribute. PubFig Dataset is used for the project

File Structure

• MATLAB Files: All matlab files contains code for GIST feature extraction
• CV_Project.ipynb: Main implementation of project.
• images: PubFig dataset of 772 images
• attribs.npy: trained weights for each attribute (Using Rank-SVM)
• attribs1.npy: trained weights for seen classes for zero shot learning