Litcius/Paper detail

Skin Melanoma Segmentation using VGG-UNet with Adam/SGD Optimizer: A Study

V. Rajinikanth, Seifedine Kadry, Robertas Damaševičius, D. Sankaran, Mazin Abed Mohammed, Shrinithi Chander

20222022 Third International Conference on Intelligent Computing Instrumentation and Control Technologies (ICICICT)28 citationsDOI

Abstract

Skin is a major sensory organs and the irregularity in the skin causes various issues. Skin cancer is one major medical emergency, and appropriate detection and treatment are essential. Clinical level skin cancer assessment is normally performed using the Dermoscopy Images (DI). This research aims to develop and implement the VGG-UNet scheme to extract and assess the abnormal section in DI with improved accuracy. This plan includes; (i) Image collection and resizing, (ii) Segmentation of skin lesion in DI, (iii) Comparing the segmented section with the Ground-Truth (GT), and (iv) Performance evaluation and validation. The merit of the proposed technique is confirmed by computing the necessary performance metrics. In this work, the performance of the proposed technique is verified with different optimizers (Adam/SGD) and the average pooling. The experimental work is executed on the ISIC2016 challenge dataset, and the merit of this scheme is verified with Jaccard, Dice, and Accuracy.

Topics & Concepts

Jaccard indexGround truthSegmentationComputer sciencePoolingArtificial intelligenceSkin cancerDiceImage segmentationPattern recognition (psychology)CancerMedicineMathematicsInternal medicineGeometryCutaneous Melanoma Detection and ManagementAI in cancer detectionNonmelanoma Skin Cancer Studies
Skin Melanoma Segmentation using VGG-UNet with Adam/SGD Optimizer: A Study | Litcius