custom YOLOv8 dataset

Dataset Overview

Dataset Name

• Development‑of‑Corrosion‑Detection‑System‑for‑Mobile‑Based‑Platform‑using‑YOLOv8

Keywords

• AI‑based corrosion detection
• Custom YOLOv8 dataset
• Early detection
• Mobile‑based platform

Dataset Content

• Image Quantity & Type: Collected 705 corrosion‑surface images, expanded to 1,398 images via augmentation.
• Dataset Split: Training set (1,200 images), Validation set (141 images), Test set (57 images).
• Annotation: Instance‑segmentation annotations using Roboflow.
• Data Augmentation: Techniques such as flipping and resizing.

Model & Methodology

• System Architecture: AI‑driven corrosion detection system based on YOLOv8, operated via a mobile interface, with Flask managing HTTP interactions and file handling.
• Instance‑Segmentation Detection: Detects both presence and boundaries of each object instance.
• Model Training: Detailed description of architecture choices and training environment.
• Evaluation Metrics: IoU (Intersection over Union), Precision, Recall.

Results & Discussion

• Performance Evaluation: Shows accuracy, recall, and video rendering time for different models (YoloV8x, YoloV8n, YoloV5x, YoloV5n).
• Model Deployment: Deploys via Flask, supporting video upload and real‑time camera detection.

Conclusion

• Successfully developed a YOLOv8‑based mobile AI corrosion detection system that outperforms YOLOv5 in both accuracy and speed, providing an effective solution for early corrosion detection.