RMNv2: Reduced Mobilenet V2 for CIFAR10

If you need an accessible version of this item, please email your request to digschol@iu.edu so that they may create one and provide it to you.
Date
2020-01
Language
English
Embargo Lift Date
Committee Members
Degree
Degree Year
Department
Grantor
Journal Title
Journal ISSN
Volume Title
Found At
IEEE
Abstract

In this paper, we developed a new architecture called Reduced Mobilenet V2 (RMNv2) for CIFAR10 dataset. The baseline architecture of our network is Mobilenet V2. RMNv2 is architecturally modified version of Mobilenet V2. The proposed model has a total number of parameters of 1.06M which is 52.2% lesser than the baseline model. The overall accuracy of RMNv2 for CIFAR10 dataset is 92.4% which is 1.9% lesser than the baseline model. The architectural modifications involve heterogeneous kernel-based convolutions, mish activation, etc. Also, we include a data augmentation technique called AutoAugment that contributes to increasing accuracy of our model. This architectural modification makes the model suitable for resource-constrained devices like embedded devices, mobile devices deployment for real-time applications like autonomous vehicles, object recognition, etc.

Description
item.page.description.tableofcontents
item.page.relation.haspart
Cite As
Ayi, M., & El-Sharkawy, M. (2020). RMNv2: Reduced Mobilenet V2 for CIFAR10. 2020 10th Annual Computing and Communication Workshop and Conference (CCWC), 0287–0292. https://doi.org/10.1109/CCWC47524.2020.9031131
ISSN
Publisher
Series/Report
Sponsorship
Major
Extent
Identifier
Relation
Journal
2020 10th Annual Computing and Communication Workshop and Conference
Source
Author
Alternative Title
Type
Conference proceedings
Number
Volume
Conference Dates
Conference Host
Conference Location
Conference Name
Conference Panel
Conference Secretariat Location
Version
Author's manuscript
Full Text Available at
This item is under embargo {{howLong}}