Object Detection Using Convolution Neural Networks
8 Pages Posted: 18 Jul 2019 Last revised: 30 Sep 2019
Date Written: May 17, 2019
Accurate and efficient object detection has been an important topic in the advancement of computer vision systems. With the sophistication of deep learning techniques, the precision of object detection has increased drastically. A huge number of visually impaired people in the world have inspired many smart solutions which use sophisticated technologies to aid them in their day to day life. This paper describes a system that aims to perform object detection with the goal of achieving high precision with acceptable real-time performance. The system is designed to help the visually impaired locate day today objects through object detection. The system is trained on CIFAR-100  dataset which consists of 100 classes where each class has 500 training samples and 100 testing samples. The total training size is 50000 samples and the testing size is 10000 samples. The experimental results are determined by using convolution neural network combined with dropout, batch normalization and data augmentation to determine which technique or combination of techniques provides the best performance for object detection. The system is then implemented with the combinations of different technique.
Keywords: Computer Vision, Object Detection, Deep Learning, Convolution Neural Network
JEL Classification: Y60
Suggested Citation: Suggested Citation