Zalo Traffic Sign Dataset

The Zalo Traffic Sign Detection 2020 Dataset presents a significant challenge yet is highly suitable for traffic sign detection applications in the diverse traffic conditions of Vietnam. Comprising 4,500 images sized at 1622x626 pixels, the dataset was curated by crawling Google Street View Map across various locations in Vietnam. The task entails detecting seven prevalent types of traffic signs: no entry, no parking/waiting, no turning, max speed, other prohibition signs, warning, and mandatory. Notably, the dataset poses considerable challenges due to the majority of objects of interest being extremely small, often less than 10x10 pixels in size. Additionally, it encompasses a wide array of scenes, ranging from urban areas to rural landscapes and from bustling city streets to secluded highways. The dataset is partitioned into two subsets: the train set, comprising 4,000 images, and the test set, comprising 500 images. Although the size of images of the Zalo dataset is large, traffic signs are very small in comparison to the whole image because the signs are very far from the camera.

Zalo 交通标志检测 2020 数据集呈现了一项重大挑战，然而对于越南多样化的交通条件下交通标志检测应用而言，该数据集却极具适用性。该数据集由 4,500 张尺寸为 1622x626 像素的图像组成，经过对越南各地 Google Street View 地图进行爬取而精心编纂。任务要求检测七种常见的交通标志类型：禁入、禁停/候车、禁转、最高车速、其他禁止标志、警告标志和强制性标志。值得注意的是，由于大部分目标对象极其微小，通常尺寸小于 10x10 像素，因此该数据集颇具挑战性。此外，它涵盖了从城市区域到乡村景观，从繁华的城市街道到偏僻的高速公路的广泛场景。数据集分为两个子集：训练集，包含 4,000 张图像，以及测试集，包含 500 张图像。尽管 Zalo 数据集的图像尺寸较大，但与整个图像相比，交通标志非常小，因为标志距离摄像头非常遥远。