top of page

What is ACID?

æ©Ÿå…·-02.png
æ©Ÿå…·-02.png
æ©Ÿå…·-07.png
æ©Ÿå…·-03.png
æ©Ÿå…·-08.png
æ©Ÿå…·-04.png
æ©Ÿå…·-09.png
æ©Ÿå…·-05.png
æ©Ÿå…·-10.png
æ©Ÿå…·-06.png
æ©Ÿå…·-11.png

Features​

  • Support object detection, instance segmentation and image captioning tasks in construction

  • 10 categories of construction machines

  • 10,000 labeled images

  • 15,767 construction machine objects

  • 19,547 captioning sentences

The Alberta Construction Image Dataset (ACID) supports three deep learning tasks: object detection, instance segmentation, and image captioning in construction, along with providing algorithm analysis results. Developed as a resource to facilitate the use and development of deep learning applications in the construction automation field, the dataset contains images collected from construction sites worldwide and is available for download.

 Dataset image examples

Construction Applications Powered by ACID

AI algorithms can automatically identify construction objects/instances/scenes in images and videos through parallel computation and graphic card implementation. However, they require large-scale datasets to mitigate the overfitting problem during the training stage. Since construction sites are often challenging to access, creating extensive datasets containing high-quality images of construction machines poses a significant challenge.

 

ACID has been developed as a construction machine dataset to facilitate the training of deep learning algorithms in construction sites. ACID supports a wide range of AI applications in construction in terms of on-site monitoring, productivity analysis, and safety management. The following videos showcase various AI applications backboned by ACID in construction.

Download and Cite ACID Paper

Two papers have been published with ASCE for ACID. If you find ACID is helpful to your work, please consider citing our papers:

 

Dataset Development and Object detection (download link)

Xiao, B., & Kang, S. C. (2021). Development of an image data set of construction machines for deep learning object detection. Journal of Computing in Civil Engineering, 35(2), 05020005.

​

Image Captioning (download link)

Xiao, B., Wang, Y., & Kang, S. C. (2022). Deep learning image captioning in construction management: a feasibility study. Journal of Construction Engineering and Management, 148(7), 04022049.

Impacts of ACID

Until now, ACID has been shared with over 400 researchers from universities/companies of 30 different countries.  

interview.png
图片3.png
图片2.png

ACID users come from diverse backgrounds and research communities, such as construction management, computer science, civil engineering, and etc. Below shows the ACID user demographic. 

ACID Statements

ACID Dataset

Alberta Construction Image Dataset (ACID) is a comprehensive image dataset for training AI models in construction automation. The development of ACID has experienced several stages and various researchers have contributed to it (see here for development roadmap of ACID). The intellectual property of ACID belongs to the ACID group, which is not affiliated to any specific university/institution. 

ACID Group

ACID group is a group of researchers from North America, Asia, and Australia (see ACID people) working together to push the boundaries of AI in Construction. The ACID group is currently taking the responsibility to maintain and develop ACID dataset. 

​

​

bottom of page