Genius-Society/tt100k
收藏资源简介:
Tsinghua–Tencent 100K (TT100K) 数据集是一个大规模的交通标志基准数据集,源自 100,000 张腾讯街景图像,包含超过 30,000 个标注实例,涵盖 221 个类别。该数据集专为现实世界中的检测和分类任务设计,具有光照、天气、视角和距离等方面的显著变化。虽然原始研究侧重于使用每类 100 个实例阈值的 45 类子集,但 Ultralytics 配置保留了所有 221 个类别,包括许多样本稀疏的类别,这使得数据集既全面又具有挑战性,适用于稳健的模型开发。
The Tsinghua–Tencent 100K (TT100K) dataset is a large-scale traffic sign benchmark dataset derived from 100,000 Tencent Street View images, containing over 30,000 annotated instances spanning 221 categories. This dataset is specifically designed for real-world detection and classification tasks, featuring significant variations in illumination, weather, viewing angles and distances. While the original research focused on a 45-class subset with a threshold of 100 instances per class, the Ultralytics configuration retains all 221 categories, including many sparsely-sampled categories, which makes the dataset both comprehensive and challenging, suitable for robust model development.




