Improving Image Search Accuracy using Deep CNN Features and KNN Matching

This study is about making it easier to find the pictures in a big collection of images. The old ways of doing this, like using words to describe the pictures or using features that people come up with, do not work well. They do not really understand what the pictures are about. A system utilizes deep learning (DL) to look at the pictures and find the important parts. Utilizing Convolutional Neural Networks (CNN’s) is to do this. The is look at the pictures. Make a list of what is in them. This list is very long. Has a lot of extra information that does not need. Utilizing Principal Component Analysis (PCA) is to make the list shorter and easier to utilize. This way, compare pictures with each other. Find the similar ones. Utilize the K-Nearest Neighbor algorithm (KNN) to do this. When tried this out, found that using PCA makes it faster to find the pictures without making any mistakes. Our system is good for collections of pictures, and it does not utilize too much computer power.