Dataset of manually classified images obtained from a construction site

A manually classified dataset of images obtained by four static cameras located around a construction site is presented. Eight object classes, typically found in a construction environment, were considered. The dataset consists of 1046 images selected from video footage by a frame extraction algorithm and txt files containing the objects' class and coordinates information. These data can be used to develop computer vision techniques in the engineering and construction fields.


Value of the Data
• This data provides 1046 manually classified images and their corresponding .txt classification files, including 8 types of objects ( Table 1 ) found in a construction site. The images were taken from four cameras placed at different points, at different moments, around a construction site. • The data presented is useful for researchers who wish to use or add these classified images to their databases, for further classification and object detection training through construction monitoring systems with computer vision. • This dataset can be used to validate a neural network of object recognition.
• This dataset contains images that are specific to construction sites, focusing on both machinery and construction personnel.

Data Description
Images of the construction site for the University Wellness Center located at Universidad de Lima were obtained by four static cameras video footage ( Fig. 1 ). A total of 1046 images were collected. This data was verified by Del Savio et al. [1] to use artificial intelligence in object detection in a construction site. Table 1 shows the classes used for manual classification and their respective images. These elements are found in the classes.txt file. As the elements are part of a Python list, the ID starts with the number 0 for element 1, and finishes with ID 7 for element 8.
These classes could be further classified in groups according to the needs of the research, particularly in cases where our granularity level is not desired, as per the example shown in Fig. 2 . This hierarchical classification proposed, however, was not applied to the dataset. Table 2 shows two examples used for the manual classification: the original images, the images during the manual classification, and the results, in .txt format, after the process. The first column presents the images obtained through the execution of the frame extraction algorithm, from the videos of the surveillance tools. The images are in jpg format, with 3840 × 2160 pix-  Table 2 Examples of selected images before, during and after classification.

Original image
Images during classification Results in .txt format

Experimental Design, Materials and Methods
This dataset was gathered using four static cameras: a bullet-type camera (Dahua Technology, 8MP Lite IR Vari-focal Bullet Network Camera) and three motorized IP PTZ cameras (Dahua Technology, 4K 48x Starlight + IR WizMind Network PTZ Camera). These cameras recorded video footage from four different points of view around the construction site ( Fig. 4 ) using the video management system DSS Express [4] . The images used were collected between November 2020 and February 2021, Table 3 shows the date and time, according to the ID of the image Table 4 . shows an average of the weather conditions during the dates of the collected images, describing the illuminance and air temperature.
A frame extraction algorithm was used to obtain the images from the video footage in jpg in a 3840 × 2160 pixels format, every 200 frames. The images went through a manual classification process with the LabelImg v.1.8.1 software [3] to identify the construction objects on site by creating quadrilaterals around the objects and assigning them a class. The results were exported in a .txt file for each image.

Ethics Statement
This research did not involve any human subjects, animal experimentation nor social media platforms.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships which have or could be perceived to have influenced the work reported in this article.