Update README.md

# Scale Noise-Robust Pavement Crack Detection Dataset

CrackBlock600K is a large-scale, high-resolution pavement crack dataset designed for noise-robust crack detection, label refinement, and segmentation research.
It consists of **600,000 overlapped 1000×1000 image patches** derived from **50,000 real-world highway images** collected in Zhejiang Province, China.

This dataset supports training, evaluation, label noise mitigation, and attention-guided crack segmentation.

------

## 📂 Dataset Structure

The dataset contains:

```
CrackBlock600K/
├── images/ # 50k grayscale 3400×2200 images
├── labels/ # 600k corresponding block-level annotation maps
└── README.md
```

### ● Label format

- Value:
- `1` → crack present
- `0` → no crack

------

## 🏗️ Dataset Construction

### 1. **Data Collection**

- Collected using a CiCS I inspection vehicle
- Fixed camera parameters (focal length 60mm)
- Driving speed: 30 km/h
- 50,000 grayscale road images (2000×3000)
- Standardized to **2200×3400**

### 2. **Annotation Protocol**

Three-stage annotation pipeline:

1. Manual annotation
2. Expert verification
3. Cross-checking for consistency

------

## 🎯 Intended Use

CrackBlock600K is designed for:

- Pavement crack segmentation
- Noise-robust training
- Label refinement research (e.g., Co-teaching, GMM clustering)
- Attention-based crack detection models
- Benchmarking U-Net/DeepCrack/SAM-assisted segmentation

It supports both **classification** (block-level) and **segmentation** tasks.

------

## 📑 License

- CC-BY-4.0

------

## 🤝 Acknowledgements

This work was supported by Shenzhen Science and Technology Program under Grant No.: CJGJZD20240729113700002, the Beijing Natural Science Foundation under Grant No.4252009 and the Fundamental Research Funds for the Central Universities under Grant No.:2023RC69.

------

README.md

@@ -1,3 +1,11 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+task_categories:
+- feature-extraction
+tags:
+- crack-detection
+- segmentation
+- road-maintenance
+size_categories:
+- 10K<n<100K
+---