Blurred Image Classification

Abstract

Without incorporating any human assistance at any stage, picture categorization is a crucial
component of image processing for computer vision and machine learning. Investigations
into high-profile crimes sometimes include the classification and identification of digital
images extracted from CCTV footage. In this research, we investigate the picture
categorization and detection of extremely pixelated, fuzzy images that are captured from
CCTV footage of cameras placed in public areas. This paper demonstrates how artificial
intelligence and deep learning principles can be used to accomplish this. To identify blurry
images, we have developed a model that combines the Tensorflow architecture with the
Convolutional Neural Network (CNN) and Sequential models. For our learning model,
60,000 photos of various object classes were collected and divided into the test dataset and
training dataset categories. In order to get the findings, a custom neural network using the
architecture of a convolutional neural network and the Keras API is used to forecast which
class a specific blurred image belongs to.