A Novel Approach for Weakly Supervised Object Detection Using Deep Learning Technique

Abstract

Computer vision has enormous potential in understanding useful representations and patterns from images and videos so that it helps to make intelligent decisions in a similar manner as humans learn from their surroundings. There has been a significant advance in the field of computer vision application: object detection. It recently came into prominence due to the emergence of deep learning techniques. Object detection is applied in wide areas of applications, like crowd detection, face detection, medical image analysis, video analysis, military applications, self-driving cars, etc. This chapter focuses on the latest advancements in object detection using deep learning techniques. These object detection techniques can be used to solve the problem of solid waste material detection so as to maximize the recycling process and automate the process of waste segregation. The well-known problem of object detection is to train the deep learning model for object classification using object-level annotations which is more time consuming and expensive. This chapter helps to provide the solution for solid waste material detection by using weakly supervised object detection. It works on image-level annotations only. This would be useful to locate recyclable items from the multiple objects present in a scene. Pre-trained deep learning models are used for feature extraction of input images which is an essential step before actual object detection tasks. Finally, promising future directions are provided for researchers.