iNaturalist Species Classification and Detection Dataset (iNat2017)
收藏资源简介:
iNaturalist Species Classification and Detection Dataset (iNat2017) 是一个大规模的图像数据集,由加州理工学院、Google、康奈尔科技和iNaturalist合作创建。该数据集包含859,000张来自全球超过5,000种不同动植物的图像,特别强调视觉上相似的物种和多样化的拍摄环境。数据集的收集过程涉及多种相机类型和图像质量,以及由多位公民科学家验证的大类不平衡。iNat2017数据集旨在推动野外数据的大规模物种分类和检测技术的发展,特别是在处理类别不平衡和细粒度分类方面的挑战。
The iNaturalist Species Classification and Detection Dataset (iNat2017) is a large-scale image dataset co-created by the California Institute of Technology, Google, Cornell Tech, and iNaturalist. It encompasses over 859,000 images representing more than 5,000 distinct animal and plant species sourced from across the globe, with a particular focus on visually similar species and diverse capture environments. The dataset was collected using multiple camera types and a spectrum of image qualities, and it features a notable class imbalance that was validated by multiple citizen scientists. The iNat2017 dataset is intended to advance the development of large-scale species classification and detection technologies leveraging field-collected data, specifically addressing the challenges of class imbalance and fine-grained species classification.

- 1The iNaturalist Species Classification and Detection Dataset加州理工学院 · 2018年



