Real time object identification using deep convolutional neural networks

The project on Real Time Object Identification presents an approach to use the concept of deep learning with the convolutional neural networks in identifying the objects present when video is given as the input. The method uses the input video to give the output with the set of identified objects su...

Full description

Saved in:
Bibliographic Details
Published in2017 International Conference on Communication and Signal Processing (ICCSP) pp. 1801 - 1805
Main Authors Sujana S., Rajeswari, Abisheck S., Sudar, Ahmed A., Tauseef, Chandran K.R., Sarath
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.04.2017
Subjects
Computer architecture
Convolutional neural networks
Detectors
Microsoft Windows
Object recognition
Real-time object identification
Real-time systems
Residual Networks
Single Shot Multibox Detector
Training
VGGNet
Online AccessGet full text

Cover

More Information
Summary:The project on Real Time Object Identification presents an approach to use the concept of deep learning with the convolutional neural networks in identifying the objects present when video is given as the input. The method uses the input video to give the output with the set of identified objects surrounded by boxes, even if the objects are of different sizes and shapes. Along with identification of the objects, the convolutional neural network works give the confidence score for each of the object. This methodology is called the Single Shot MultiBox Detector (SSD). In this project, we are replacing the VGG Net with Residual Networks in the architecture to increase the computational speed.
DOI:10.1109/ICCSP.2017.8286705