A dissected dataset for single and double wildlife fences in South Africa

The dataset consists of 211 images collected using a standalone camera and a drone, stored in jpeg format with a height and width of 512 and 512 pixels, respectively. The image dataset contains 105 single fences and 106 double fences. The drone camera images included 26 single fences and 26 double fence images. In contrast, the standalone camera has 159 images with 79 single and 80 double fences. We labeled the datasets as either still or aerial datasets. The still dataset contains only images that used a standalone camera; the drone captured aerial images of wildlife fences and scenes from the sky. The images can be used to develop deep learning algorithms that classify whether a wildlife fence is single or double. This study uses wildlife electric fences for both single and double. When machine learning algorithms detect electric fences, dangerous animals, such as lions, are in the game reserve, and road users expect to take extra precautions while using the road. In addition, the double fence ensures high double protection in case of failure of the inner fence to protect wildlife from escaping.

本数据集包含211张图像，均通过独立式相机（standalone camera）与无人机（drone）采集，存储格式为JPEG（JPEG），分辨率均为512×512像素。该图像数据集包含105张单围栏与106张双围栏图像。其中无人机相机采集的图像包含26张单围栏与26张双围栏图像；相较之下，独立式相机采集的159张图像中包含79张单围栏与80张双围栏图像。本数据集被划分为静态数据集（still dataset）与航拍数据集（aerial dataset）两类：静态数据集仅包含独立式相机采集的图像；无人机采集的航拍图像涵盖了保护区内的野生动物围栏与空中场景。本数据集可用于开发用于分类野生动物围栏为单围栏或双围栏类型的深度学习算法（deep learning algorithms）。本研究涉及的单、双围栏均为带电野生动物围栏。当机器学习算法（machine learning algorithms）检测到带电围栏时，表明保护区内存在狮子等危险动物，道路使用者途经该路段时需采取额外防范措施。此外，双围栏可在内层围栏失效时提供双重防护，有效防止野生动物逃逸。