Out of Context Object Detection

Date
2022
Editor
Journal Title
Volume
Issue
Journal Title
Journal ISSN
Volume Title
Abstract
Image segmentation is a sub-domain of computer vision and digital image processing which aims at grouping similar regions or segments of an image under their respective class labels. Image segmentation forms the basis for object detection. Object detection is fairly common in computer vision as it is beneficial in fields of Robotics and object recognition. However, there is a lot of scope on working with Out of Context Objects Detection in images which is not being done. Therefore, in this project, we chose Detectron2, a segmentation model, and a highly annotated contextual dataset such as the COCO train17 dataset and used it to identify objects in contextual image. We implemented the image segmentation model and results show that the model was successful in identifying and segmenting different classes and objects from scenes containing many different types of objects. The object detection model is used to create a dataset of objects commonly found in contextual scenarios and the dataset is used to train the Word2Vec model. The trained Word2Vec holds the capability to identify any out of context objects from a list of objects presented to it.
Description
Keywords
Citation
Department Name
Electrical and Computer Engineering
Publisher
North South University
Printed Thesis
DOI
ISSN
ISBN