Kinzaaa/nut_defect_detection

Nut Defect Classification

Synthetic Industrial Quality Inspection Dataset

Nut Defect Classification (Synthetic Dataset)

This dataset is a synthetic collection of industrial nut images designed for image classification tasks, specifically focusing on defect detection in manufacturing pipelines. It serves as a benchmark and training resource for computer vision algorithms used in quality assurance.

Dataset Structure

The dataset contains a total of 261 images categorized into two classes:

• defect: Images representing industrial nuts with defects (e.g., structural, surface flaws).
• non_defect: Images representing normal, non-defective nuts.

The dataset includes a metadata.csv mapping image paths to their respective labels, as well as a dataset-metadata.json describing the dataset details.

Directory Layout

├── README.md
├── dataset-metadata.json
├── metadata.csv
├── synthetic_defect/
│   ├── synth_defect_1.png
│   └── ...
└── synthetic_non_defect/
    ├── synth_non_defect_1.png
    └── ...

Data Fields

The metadata.csv contains the following fields:

• file_name: Path to the image file relative to the root directory (e.g. synthetic_defect/synth_defect_1.png).
• label: Class label (defect or non_defect).

Use Cases

• Quality Control & Automation: Training models to detect defect products on assembly lines.
• Anomaly Detection: Evaluating unsupervised or semi-supervised anomaly detection methods.
• Synthetic Data Research: Analyzing the transferability of synthetic datasets to real-world scenarios.

Licensing

Licensed under the MIT License.