Project Title

Downsampling select representative dataset

Description

Pipeline of select representative image dataset from a large dataset using 2 different approach: unsupervised(VGG) and traditional CV(ORB).

Pipeline structure:

• Feature extraction using traditional approach(ORB) -> Bag of Feature on image features with k-means -> PCA dimention reduction -> k-means clustering
• Feature extraction using CNN approach(VGG) -> PCA dimention reduction -> k-means clustering

Getting Started

Dependencies

• Python 3
• Tensorflow

Installing

• Pull from the github page with the link from clicking 'Code' then clone
• pip install all dependency packages

Authors

• Moya Zhu (uni: mz2886) MSDS 22' Worked with Fairvision research group