Fast and Lightweight Object Detection Network: Detection and Recognition on Resource Constrained Devices
Open Access
- 5 February 2018
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 6, 8714-8724
Abstract
The intrinsic ability of humans to rapidly detect, differentiate and classify objects allows us to make quick decisions in regards to what we see. Several appliances can make use of a fast and lightweight automated object detection for images or videos. Throughout the last 5 years, the technology industry has constantly introduced computational and hardware solutions, such as devices with impressive processing and storage capabilities. However, object detection methods usually require either high processing power or large storage availability, making it hard for resource constrained devices to perform the detection in real-time without a connection to a powerful server. The model presented in this paper requires only 95 megabytes of storage and took 113ms in average per image running on a laptop CPU, making it suitable for standalone devices that can be used on the go.Keywords
This publication has 35 references indexed in Scilit:
- Regionlets for Generic Object DetectionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
- Big Data Deep Learning: Challenges and PerspectivesIEEE Access, 2014
- Practical Recommendations for Gradient-Based Training of Deep Architectures2012
- Energy-efficient algorithmsCommunications of the ACM, 2010
- Category Independent Object Proposals2010
- A performance study of general-purpose applications on graphics processors using CUDAJournal of Parallel and Distributed Computing, 2008
- LabelMe: A Database and Web-Based Tool for Image AnnotationInternational Journal of Computer Vision, 2007
- Application of support vector machines in financial time series forecastingOmega, 2001
- Learning algorithm for nonlinear support vectormachines suited for digital VLSIElectronics Letters, 1999
- Learning representations by back-propagating errorsNature, 1986