Problem Statement : To design an Image Search Engine application which recieves a product image and recommends similar products.
We will make use of ResNet-50 architecture. Residual Network (ResNet) is a supervised deep learning model used for computer vision applications. It is a Convolutional Neural Network (CNN) architecture designed to support hundreds or thousands of convolutional layers. Find the architecture below.
Our dataset contains 6000 images, so we will have 6000 feature vector from ResNet-50 model and 6000 filenames corresponding to each feature vector.
Once a test image is inputted, the model gives a 2048 output vector. We already have a vector space of training images. So once the test image vector arrives at the space, the decision will be taken based on the Euclidean distance calculation using KNN.
