AGROINSECT: A dataset for identifying agricultural insects and pests of interest.

The AgroInsect dataset comprises 1,510 images and 1,654 annotations, unevenly distributed across classes, reflecting the natural variability inherent to field collections and the public data sources (iNaturalist) used in its construction. In certain classes, the number of annotations exceeds the number of images, as some photographs contain multiple individuals, with each identified insect treated as an independent annotation.
The dataset is organized according to the widely adopted YOLO format, in which images and their corresponding annotation files are stored separately. Annotations were created using the Label Studio platform and follow the YOLO standard, where each line contains the class identifier and the normalized coordinates of the corresponding bounding box. This structure enables the direct use of the dataset for training and evaluating deep learning–based object detection models without additional preprocessing. The AgroInsect dataset was designed to support the development, evaluation, and benchmarking of machine learning and deep learning models, with direct applications in insect classification and agricultural pest detection. Its relevance for these tasks lies in the composition of the images, which represent realistic field conditions characterized by variations in illumination, occlusions, different insect developmental stages, and complex backgrounds, thus posing significant challenges to computer vision algorithms. The practical utility of AgroInsect has been demonstrated in studies that employed the dataset to evaluate the performance of state-of-the-art object detection architectures, including models from the YOLO family, their optimized variants for embedded and edge computing environments using TFLite, as well as the Detectron2 framework. Finally, the dataset includes four insect classes of high agricultural importance—Diabrotica speciosa, Dalbulus maidis, Diceraeus spp., and Spodoptera frugiperda—selected due to their significant economic impact on soybean and maize production systems in the state of Mato Grosso and other agricultural regions of Brazil. In this context, AgroInsect represents a strategic resource for applied research in digital agriculture, contributing to the development of artificial intelligence–based solutions for pest monitoring and the promotion of sustainable agricultural practices.