Welcome to the Surface Crack Classifier project, a sophisticated computer vision model designed to address the critical task of classifying surface images into two distinct categories: surfaces with cracks and those without. This project is born out of the necessity to enhance infrastructure monitoring, material inspection, and various applications where the identification of surface defects is paramount. It is powered by MobileNet transfer learning, excels in precise classification of surface images into cracked and non-cracked categories. Tailored for infrastructure monitoring, quality control, and beyond.
MobileNet Transfer Learning: Leveraging MobileNet architecture for efficient and effective model training. High Accuracy: Achieves remarkable results with 99.8% train accuracy and 99.6% validation accuracy. Versatile Application: Ideal for civil engineering, construction, and quality assurance. Easy Integration: Seamlessly integrates into existing workflows with support for popular deep learning frameworks.
Data Preparation: Organize images into 'Cracked' and 'Non-Cracked' folders. Model Training: Utilize MobileNet transfer learning for swift and accurate training with exceptional accuracy. Inference: Deploy the model for real-time surface defect detection.
We welcome contributions and suggestions. Let's build a powerful and accessible Surface Crack Classifier together.
This dataset is taken from the website Mendeley Data http://dx.doi.org/10.17632/5y9wdsg2zt.2, contributed by Çağlar Fırat Özgenel.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent