A YOLOv8-based Model for Precise Corrosion Segmentation in Industrial Imagery

• Published in: 2024 3rd International Conference on Artificial Intelligence For Internet of Things (AIIoT) pp. 1 - 6
• Main Authors: Yuvaneswaren, R S, Prakash, S Tarun, Sudha, K.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 03.05.2024
• Subjects: Cloud computing; Corrosion; Corrosion Detection; Deep learning; Image Analysis; Image segmentation; Neural Network; Neural networks; Pipelines; Roboflow; Virtual machining; WandB; YOLOv8
• DOI: 10.1109/AIIoT58432.2024.10574659

Corrosion, the ongoing degradation of materials caused by chemical reactions with their surroundings, poses a continuous and expensive obstacle across various industrial domains. From bridges and pipelines to machinery and storage tanks, corrosion poses a substantial threat to infrastructure integrity, safety, and longevity. Timely detection and precise localization of corrosion areas are vital for proactive maintenance, reducing downtime, and ensuring the longevity of critical assets. Traditionally, corrosion inspection has relied heavily on manual, time-intensive methods, which are often limited in accuracy and efficiency. However, the advent of deep learning, a subfield of artificial intelligence, has revolutionized the domain of image analysis, offering unparalleled potential for automated and precise corrosion detection.This project undertakes a transformative journey in leveraging the capabilities of deep learning, particularly focusing on computer vision, to achieve precise and efficient corrosion segmentation in industrial imagery. The primary objective is to employ cutting-edge deep learning models, with a specific emphasis on the YOLOv8 architecture, to create a robust solution capable of accurately identifying and outlining corrosion regions within intricate industrial scenes. Furthermore, the project explores the potential of utilizing cloud-based virtual machines, specifically the WandB Cloud offering, as a cost-effective alternative to conventional solutions such as Azure for model deployment and monitoring. Noteworthy attributes of YOLOv8 include its distinctive single-pass detection methodology, where the entire video frame undergoes processing in a single forward pass through the neural network, eliminating the need for intricate post-processing and enhancing detection speed.