CropAndWeed Dataset

CropAndWeed数据集专注于对74种相关作物和杂草物种的细粒度识别，并着重于数据可变性。在现实世界的农业站点和稀有杂草类型的专门栽培的室外地块的8k以上高分辨率图像中，为大约112k个实例提供了标记的边界框，语义面具和茎位置的注释。此外，每个样品都富含有关环境条件的元注释。

The CropAndWeed dataset focuses on fine-grained recognition of 74 relevant crop and weed species, with a particular emphasis on data variability. It includes over 8,000 high-resolution images captured from real-world agricultural sites and specially cultivated outdoor plots for rare weed types, with annotated bounding boxes, semantic masks and stem position annotations for approximately 112,000 instances. In addition, each sample is enriched with meta-annotations related to environmental conditions.

搜集汇总

背景与挑战

背景概述

CropAndWeed数据集是一个专注于74种作物和杂草细粒度识别的数据集，包含8k以上高分辨率图像和112k个实例的注释信息，由AIT Austrian Institute of Technology于2023年发布。