Tolerance to spittlebugs (Hemiptera: Cercopidae) in Urochloa spp. and Megathyrsus maximus grasses

This dataset comprises 8,318 images including Urochloa spp. and Megathyrsus maximus experimental units. The dataset is a resource for validating the current plant damage quantification technique for repeatability, and for training machine learning algorithms to identify, classify or quantify plant damage caused by biotic and abiotic stress with similar symptoms. Methodology: These data were obtained from the assessment of 8 populations of the Urochloa and Megathyrsus maximus breeding programs in 15 no-choice tests. Each test had a row and column number to analyze the statistical data as a spatial design. The images were acquired prior to the infestation, and 35 days after the infestation for nymphs and prior to the infestation, 7 days after the infestation and 14 days after the infestations for adults. Each experimental unit was placed in a white, enclosed chamber (dimensions: 1x1x1 m), with a strip of LED day white lights (6000 k) for consistent illumination. Images were captured using a Canon 90D and a NIKON D7500 reflex cameras with the following set up: manual mode, focus mode AF-A single point, white balance set to 0.0 in the fluorescent mode, ISO speed set to 100, shutter speed set to 1/50s, and aperture set to 5.6. The images included in the dataset are in JPEG format, the metadata for each image was compiled in a table after assessing the resistance to damage through visual and image-processing based methods. The files were organized in a folder-based image classification format (sometimes known as ImageNet) compatible with the one required by computer vision classification models.

本数据集共收录8318张图像，涵盖臂形草属（Urochloa spp.）与大黍（Megathyrsus maximus）的试验单元。该数据集可用于验证当前植物损伤量化技术的重复性，亦可作为训练机器学习算法的资源，以实现对具有相似症状的生物胁迫与非生物胁迫所致植物损伤的识别、分类或量化。
研究方法：本数据集源自对15项无选择试验中8个臂形草属及大黍育种群体的评估。每项试验均设置行号与列号，以空间设计范式开展统计数据分析。
图像采集时序如下：针对若虫，分别在侵染前、侵染后35天进行图像采集；针对成虫，则在侵染前、侵染后7天及侵染后14天完成图像采集。
每个试验单元均被置于1×1×1米的白色密闭舱中，舱内配备6000K的LED日光白光条以保证光照一致性。图像采集采用佳能90D（Canon 90D）与尼康D7500（NIKON D7500）单反相机，拍摄参数设置为：手动拍摄模式、AF-A单点对焦模式、荧光白平衡模式下白平衡值设为0.0、ISO感光度设为100、快门速度设为1/50秒、光圈值设为f/5.6。
数据集内所有图像均采用JPEG（JPEG）格式存储。通过目视与图像处理方法评估植株抗损伤性后，将每张图像的元数据整理至表格中。数据集采用基于文件夹的图像分类格式（即常称的ImageNet格式），可直接兼容计算机视觉分类模型的输入要求。