Image Recognition and Similarity Retrieval with Convolutional Neural Networks

Image is a form of representation of data that is presented with visual beauty and has many patterns in it, in a database of irregular image datasets it is enough to make us humans who like structured, neat, and high readability data sets Therefore, the role of machine learning in helping to search for the same image will greatly help humans both in implementing search engines, tidying up image data structures automatically, etc. This research will focus on creating the same 5-image search system based on input from the user using the CNN algorithm & pretrained models such as ResNet with the expected results of model accuracy ranging from 80-95%. This was carried out using the CIFAR-10 dataset as a source. The CIFAR-10 dataset is a medium-scale image data set with a total of 60000 images, there are 10 categories in it, meaning each category contains 6000 images with the size of each image being 32x32. In general, a dataset has different image sizes, so this research study will first do an overview about whether the data for each image is enough with a size of 32x32.