Uddin, JiaNiloy, Wahidul HasanOrnab, Mostafa KamalSaha, Saurav2021-03-032021-03-0320192019-08ID 18341009ID 1824120ID 13101148http://hdl.handle.net/10361/14287Cataloged from PDF version of thesis.Includes bibliographical references (pages 28-30).This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2019.In a densely populated country like Bangladesh, fire accidents have become a fre- quent disaster that primarily be formed as a consequence of unconsciousness among the people. Therefore, detection of smoke, is a must in order to have an earlier cau- tion before the damages caused by fire. Thereby, in this paper, we have approached a deep convolutional neural network in the identification of smoke from images by using the process of image processing. The detection of smoke images recognized as a difficult task for having of a larger differentiation in textures, colors and structures. In competing with the challenges of detecting smoke, the model has developed with the help of the methodology of image processing and computer vision, through the deep convolutional neural network in the identification of smoke images. We have succeeded to gain the accuracy in a sufficient ratio. Using the model of Deep CNN, \VGG-19" and \Inception-v3" we have gained the accuracy of 82.33% and 84.67%. Moreover, for reducing the overfitting problem, we have structured an increasing amount of training data sets through the data augmentation techniques. Thus, the Deep Convolutional Neural Network has been utilized to perform in a more accurate way by gathering the accuracy in a more preferable way in the procedure of smoke detection.30 pages.enBrac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.Deep convolutional neural networkComputer visionVGG-19Inception- v3Smoke detectionThesis