Published December 12, 2022 | Version 1.0
Photo Open

Image Dataset for Object Detection of Small Size Construction Tools

  • 1. Korea Expressway Corporation
  • 2. Inha University

Description

 This is an image dataset established as input data for object detection model of small-sized construction tools. In the dataset, there are 12 classes of target tools  (bucket, cutter, drill, grinder, hammer, knife, saw, shovel, spanner, tacker, trowel, and wrench) which are typically used at indoor construction sites. 25,084 sets of image and the corresponding label data have been established and shared. 

 The diversity of objects in the images of the 12 small tools was considered by photographing tools of various shapes, sizes, and colors. In addition, to improve the model performance, images were also captured with various changes (e.g., image resolution, occlusion, lighting, and background). Among the 25,084 images in the dataset, 6,258 (25%) were obtained from the actual construction site. 

 Object annotations in each image were done by bounding boxes and were saved into a text file. The coordinates of the bounding box have the form of (Class, Center X, Center Y, Width, Height). Class refers to one of 12 construction tool types. Center X and Center Y are the center coordinates of the bounding box for an object from an image when the resolution of the image has min-max normalized. Width and Height are the width and height of the bounding box for an object, respectively, also from the image with the min-max normalized resolution.

 

The peer-reviewed publication for this dataset has now been published in " KSCE Journal of Civil Engineering" a Springer journal as follows:

* Lee, K., Jeon, C., and Shin, D. (2023, In press) "Small Tool Image Database and Object Detection Approach for Indoor Construction Site Safety" KSCE Journal of Civil Engineering. DOI: https://doi.org/10.1007/s12205-022-1011-7

Please cite this reference when using the dataset.

Notes

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No.NRF - 2019R1A2C1088824)

Files

DATA1.zip

Files (109.5 GB)

Name Size Download all
md5:e87dec52ce299be886ae733a9ee496dd
18.9 GB Preview Download
md5:765ac6da3420841673fdc181b01f39ad
26.3 GB Preview Download
md5:7fd534712155229800adf3bc825e7fc8
40.8 GB Preview Download
md5:d80cb8715b1107a762551ee2d559acf5
23.5 GB Preview Download