Lim Min Hui, Chan Hiang Hao, Ong Song-Quan
Institute for Tropical Biology and Conservation, Universiti Malaysia Sabah, Jalan UMS, 88400 Kota Kinabalu, Sabah, Malaysia.
Agrofog - Agro Technic Pte Ltd, 32 Ang Mo Kio Ind Park 2, #01-01/02 Sing Industrial Complex, 569510, Singapore.
Data Brief. 2025 May 16;60:111673. doi: 10.1016/j.dib.2025.111673. eCollection 2025 Jun.
A large image dataset with the aim of developing an insect recognition algorithm like YOLO. The dataset contains more than 25,000 annotations on the taxonomy of urban insects according to their order and the localization of the insect (as a bounding box) on a scanned image. This annotated image dataset of flying insects was collected using UV light traps placed in food warehouses, manufacturers and grocery stores in urban environments. The traps, equipped with UVA lamps (365 nm), captured a variety of insect species on sticky cards over 7-10 days. The sticky traps with all captured insects were used to create high-resolution scanned images (1200 dpi, 48-bit colour), with the resolution preserving fine morphological details of the insect, such as the antenna. To annotate the dataset for computer vision and deep learning models with detection tasks, annotation was performed using CVAT, with bounding boxes labelled by entomology experts at the order level. The dataset was intended to serve as a dataset for computer scientists or entomologists to compare the performance of deep learning models that can be used to build an automatic detection system for urban insect diversity or pest control studies.
一个大型图像数据集，旨在开发类似YOLO的昆虫识别算法。该数据集包含超过25000条关于城市昆虫分类的注释，这些注释根据昆虫的目以及昆虫在扫描图像上的定位（作为边界框）。这个带注释的飞行昆虫图像数据集是通过放置在城市环境中的食品仓库、工厂和杂货店的紫外线诱捕器收集的。这些诱捕器配备了UVA灯（365纳米），在7到10天的时间里在粘性卡片上捕获了各种昆虫物种。带有所有捕获昆虫的粘性诱捕器被用于创建高分辨率扫描图像（1200 dpi，48位颜色），该分辨率保留了昆虫的精细形态细节，如触角。为了用检测任务为计算机视觉和深度学习模型注释数据集，使用CVAT进行注释，边界框由昆虫学专家在目级别进行标记。该数据集旨在作为计算机科学家或昆虫学家的数据集，以比较可用于构建城市昆虫多样性自动检测系统或害虫控制研究的深度学习模型的性能。
