Datasets:
image
imagewidth (px) 256
3.84k
|
---|
Defect Spectrum Dataset
Welcome to the Defect Spectrum dataset repository. This comprehensive benchmark is a granular collection of large-scale defect datasets with rich semantics, designed to push the frontier of industrial defect inspection research and applications.
Paper: https://huggingface.co/papers/2310.17316
Github repository: https://github.com/EnVision-Research/Defect_Spectrum
Overview
Defect inspection is a critical component within the closed-loop manufacturing system. To facilitate advanced research and development in this domain, we introduce the Defect Spectrum dataset. It offers precise, semantics-abundant, and large-scale annotations for a wide range of industrial defects. This dataset is an enhancement over existing benchmarks, providing refined annotations and introducing detailed semantic layers, allowing for the distinction between multiple defect types within a single image.
Features
- Semantics-Abundant Annotations: Each defect is meticulously labeled, not just at the pixel level but with rich contextual information, providing insights into the defect type and implications.
- High Precision: Annotations are refined by experts to capture even the subtlest of defects, ensuring high precision.
- Large-Scale Data: Building on four key industrial benchmarks, Defect Spectrum stands out with its extensive coverage and depth.
- Incorporates Descriptive Captions: To bridge the gap towards Vision Language Models (VLMs), each sample is accompanied by a descriptive caption.
Directory Structure
DefectSpectrum/
βββ DS-MVTec/
β βββ bottle/
β β βββ image/ # Original images of the bottle category
β β βββ caption/ # Descriptive captions of the bottle category
β β βββ mask/ # Single channel defect masks for the bottle category
β β βββ rgb_mask/ # Colored defect masks for better visualization
β βββ cable/
β β βββ image/ # Original images of the cable category
β β βββ caption/ # Descriptive captions of the cable category
β β βββ mask/ # Single channel defect masks for the cable category
β β βββ rgb_mask/ # Colored defect masks for better visualization
β βββ ...
βββ DS-VISION/
β βββ ...
βββ DS-DAGM/
β βββ ...
βββ DS-Cotton-Fabric/
β βββ ...
To-Do List
- Task 1: Release DS-MVTec image-mask pairs.
- Task 2: Release DS-VISION, DS-DAGM, and DS-Cotton-Fabric image-mask pairs.
- Task 3: Release captions.
- Task 4: Release selected synthetic data.
license: mit
- Downloads last month
- 2,085