International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 05 | May 2024
www.irjet.net
p-ISSN: 2395-0072
Development of Smart Alerting System using Real Time Object Detection with Deep Learning Satish Dekka1, Karanam Prameela2, Padala Mohan Anantha Rama Reddy3, Manapuram Sai Charan4, Kadali Varahala Bhagya Sri5 1 Associate Professor, Dept. of Computer Science Engineering, Lendi Institute of Engineering and Technology
Vizianagaram, India 2-5B.Tech Final Year, Computer Science Engineering, Lendi Institute of Engineering and Technology Vizianagaram,
India ---------------------------------------------------------------------***-------------------------------------------------------------------Abstract—Deep learning is currently the mainstream method of object detection. This project introduces a novel approach that enhances real-time object detection by integrating the You Only Look Once (YOLO) algorithm with Fast SMS capabilities. Our proposed system leverages the strengths of YOLO to handle complex scenarios such as occlusion, deformation, and small object sizes. The system's architecture involves the integration of Deep learning with Fast SMS Services, enabling real-time detection and interaction with the physical environment. This fusion of advanced deep learning with Fast SMS offers a comprehensive solution for efficient and responsive object detection in dynamic environments. Trained on nearly 40 classes, the trained model uses Darknet for class implementation. As a real-time object detector, the system detects objects while the webcam is on and sends messages to mobile phones using Fast SMS Service. The integration of both deep learning and Fast SMS Service enables efficient and rapid identification of objects in streaming video feeds. The system leverages the power of YOLO's single-pass architecture for simultaneous detection of multiple objects with high accuracy. The detected object information is seamlessly communicated to different devices, facilitating quick decision making and response in various applications such as smart surveillance, automated monitoring, and real-time analytics.
development of Convolutional Neural Network (CNN) algorithms, starting with the Neocognitron-inspired algorithm in the 1990s by Yann LeCun et al., marked a significant milestone in object detection. This was further advanced by AlexNet, which achieved a groundbreaking victory in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012. These CNN algorithms have since been instrumental in offering solutions to object detection problems using various approaches. To enhance the accuracy and speed of the recognition, subsequent optimization-focused algorithms have been introduced. Notable examples include VGGNet, GoogLeNet, and Deep Residual Learning (ResNet), which have been developed over the years. These algorithms have played a crucial role in advancing object detection capabilities, pushing the boundaries of what is achievable in terms of accuracy and efficiency. Computer vision and visual processing of images and tones have traditionally been significant for large organizations with the resources to invest in advanced technology. However, the advent of affordable computers and massproduced devices like YOLO (You Only Look Once) has democratized these capabilities. YOLO, originally developed for commercial and academic purposes, now enables enthusiasts worldwide to develop software and real-time embedded programs that can process images using current hardware standards. The applications of software for object detection and computer vision are diverse and abundant. Examples range from tracking visitors in shopping malls and face recognition systems to monitoring equipment, assisting robots, analyzing product labels, and deep learning projects.
Keywords— Deep Learning, Object Detection, You Only Look Once (YOLO) Algorithm, Fast SMS, Smart Surveillance, Real-Time Analytics.
I. Introduction Over the years, various strategies have been proposed to address the challenge of object identification, with a focus on multi-stage solutions. These approaches typically involve stages such as recognition, classification, localization, and object detection. However, despite technological advancements, these techniques have encountered challenges related to output accuracy, resource costs, processing speed, and complexity. The
© 2024, IRJET
|
Impact Factor value: 8.226
This Bachelor's Thesis aims to create a database on embedded systems, real-time image processing, and visual effects to explore further layers of complexity and
|
ISO 9001:2008 Certified Journal
|
Page 467