Paper Title :Transfer Learning or Custom CNN Network for Special Cases of Image Classification
Author :Acharya Sudipta, Acharya Patatri
Article Citation :Acharya Sudipta ,Acharya Patatri ,
(2024 ) " Transfer Learning or Custom CNN Network for Special Cases of Image Classification " ,
International Journal of Advance Computational Engineering and Networking (IJACEN) ,
pp. 29-33,
Volume-12,Issue-4
Abstract : Convolutional Neural Networks (CNNs) have outperformed the traditional fully connected ANN (Artificial
Neural Network) in the field of image analytics including image classification, localization etc. In this paper we have
investigated if always a transfer learning based very deep CNN such as Resnet 50 or MobileNetV2 are the best choices for
all kind of image classification tasks or there can be a room of optimization and a hand-crafted network works better. It has
been found that while a 50 CNN layer network such as Resnet50 gave only 34% accuracy, a custom 11 CNN layered
networkgave 59% accuracy and most importantly another custom network with 6 CNN layers provided an accuracy of 66%,
whereas, the MobileNetV2network well known for its capability of processing low resolution images gave an accuracy of
only38%. This clearly shows that not always a very deep neural network trained on a huge dataset like Image Net performs
best. Moreover, we have also investigated the effect of batch size on accuracy for a few of these networks
Type : Research paper
Published : Volume-12,Issue-4
DOIONLINE NO - IJACEN-IRAJ-DOIONLINE-20708
View Here
Copyright: © Institute of Research and Journals
|
![](images/permissions-active.png) |
| |
![](images/pdf_icon.png) |
PDF |
| |
Viewed - 5 |
| |
Published on 2024-07-10 |
|