IJCRAR is now DOI (CrossRef) registered Research Journal. The DOIs are assigned to all published IJCRAR Articles.

INDEXED IN INDEX COPERNICUS - ICI Journals Master List 2016 - IJCRAR--ICV 2016: 81.15 For more details click here

Abstract            Volume:10  Issue-5  Year-2022         Original Research Articles


Online ISSN : 2347 - 3215
Issues : 12 per year
Publisher : Excellent Publishers
Email : editorijcret@gmail.com

Cloud-Based Transfer Learning Framework for Faster and Accurate Skin Lesion Diagnosis
Raj Kumar Gudivaka1, Rajya Lakshmi Gudivaka2, Sri Harsha Grandhi3, Dinesh Kumar Reddy Basani4, Basava Ramanjaneyulu Gudivaka5 and Aravindhan Kurunthachalam6*
1Birlasoft Limited, Telangana, India 2Jawaharlal Nehru Technological University, Kakinada, Andhra Pradesh, India 3Intel, Folsom, California, USA 4CGI, British Columbia, Canada 5JP Morgan Services India Private Limited, Telangana, India 6Assistant professor, SNS College of Technology, Coimbatore, Tamil Nadu, India
*Corresponding author
Abstract:

Deep learning systems for detecting skin cancers have been discovered as a veritable tool with which to help in achieving accuracy and efficiency in the detection of skin cancers. In the present work, a deep learning architecture for skin cancer classification is proposed, using DenseNet-121 for hierarchical feature extraction and making it suitable for real-time inference using Vertex AI Endpoints. The dataset used for evaluation originates from the ISIC Archive and consists of images of benign and malignant skin lesions. The image pre-processing steps include Gaussian Filtering and contrast-limited adaptive Histogram Equalization for improving quality and contrast. It was reported that the proposed model achieved a validation accuracy of 97.43%, precision of 85.24%, and recall of 93.44%, compromising on an F1-score of 85.88%. However, the ROC curve analysis shows a weak discrimination, with an AUC of 0.5626, thus indicating an immediate need for an improvement of the process.

Keywords: Contrast-Limited Adaptive Histogram Equalization (CLAHE), Convolutional Neural Network (CNN), Convolutional Layer (Conv Layer), Fully Connected Layer (FC).
Download this article as Download

How to cite this article:

Raj Kumar Gudivaka, Rajya Lakshmi Gudivaka, Sri Harsha Grandhi, Dinesh Kumar Reddy Basani, Basava Ramanjaneyulu Gudivaka and Aravindhan Kurunthachalam. 2022. Cloud-Based Transfer Learning Framework for Faster and Accurate Skin Lesion Diagnosis.Int.J.Curr.Res.Aca.Rev. 10(5): 131-140
doi: https://doi.org/10.20546/ijcrar.2022.1005.012
Copyright: This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike license.


Quick Navigation


Quick Navigation