Tsinghua-Tencent 100K Tutorial

尽管在交通标志检测和分类领域取得了可喜的成果，但很少有作品为现实世界的图像同时提供这两个任务的解决方案。我们为这个问题做出了两个贡献。首先，我们从100000腾讯街景全景图创建了一个大型交通标志基准，超越了以前的基准。它提供包含30000交通标志实例的100000图像。这些图像涵盖了照度和天气条件的大变化。基准中的每个交通标志都带有类标签，其边界框和像素掩码。我们把这个基准叫清华-腾讯100K

Although promising results have been achieved in the field of traffic sign detection and classification, few studies have offered solutions for both tasks simultaneously in real-world images. We present two contributions to address this issue. First, we develop a large-scale traffic sign benchmark derived from 100,000 Tencent Street View panoramas, which outperforms prior benchmarks. It comprises 100,000 images containing 30,000 traffic sign instances. These images cover extensive variations in illumination and weather conditions. Each traffic sign in the benchmark is annotated with class labels, bounding boxes, and pixel-level masks. We name this benchmark Tsinghua-Tencent 100K.