XZ-YAN / CIS-Dataset

The Construction Instance Segmentation (CIS) Dataset

CIS-Dataset

Introduction

The Construction Instance Segmentation (CIS) dataset contains 50k images with more than 83k annotated instances.

Construction Object Categories

• 2 categories of workers: workers wearing & not wearing safety helmets.
• 1 categories of materials: precast components (PCs).
• 7 categories of machines: PC delivery trucks, dump trucks, concrete mixer trucks, excavators, rollers, dozers & wheel loaders.

Download the CIS Dataset

• Download link
• The dataset is compressed for ease of download. To decode the compressed dataset, move the downloaded "dataset" folder into the root directory of this repo and run:
  $ python CIS_decoder.py train 5
$ python CIS_decoder.py val 1
$ python CIS_decoder.py test 4

Annotation Visualization

• To visualize the dataset annotations, run:
  $ python CIS_annotations_visualizer.py
• CIS annotation samples:

Dataset Statistics

• To visualize the dataset statistics, run:
  $ python CIS_statistics_visualizer.py
• CIS statistics:
  

License

• The CIS dataset is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License to promote the open use of the dataset and future improvements.
• Without permission, the CIS dataset should only be used for non-commercial scientific research purposes.

Citing the CIS Dataset

If you find this repo useful in your research, please consider citing:

• Xuzhong Yan, Hong Zhang, Yefei Wu, Chen Lin, Shengwei Liu, Construction Instance Segmentation (CIS) Dataset for Deep Learning-Based Computer Vision, Automation in Construction. 156 (2023) 105083, https://doi.org/10.1016/j.autcon.2023.105083.