Investigations on object detection And labeling using deep learning Techniques

Abstract

The primary objective of computer vision is to use computers to emulate human vision, including learning and being able to make inferences, and take actions based on visual inputs. One of the areas of computer vision is image analysis and it is also called image understanding. Image analysis is distinguished from other types of image processing like image enhancement, image transform, and image restoration where the output is projected as another separate image. In image analysis, the output is seen on the same image itself. In image analysis, image data or information is extracted from the same image and is further described or interpreted. Scene parser, one of the techniques involved in image analysis, classifies each pixel of an image into one of several predefined object classes.Scene labeling is known as scene parsing, which involves the sceneunderstanding task that assigns a label to each pixel. Scene labeling includeslabeling each pixel based on the category to which the object belongs. Thedependency of the pixel category may include either relatively short-rangeinformation or very long-range information or both. Labeling demandscontextual information because the labels tend to be dependent across pixels.Every image consists of information that is required to label pixels at several levels. In the process of identifying the pixel class, the color and the texture are sometimes sufficient.Three methods are proposed to address the problem of scene labeling.The first method utilizes the concept of labeling with Transform domain byconsidering the various textures including regular, irregular, and stochasticones. newline