TAR-1：Urban Traffic Objects Dataset

We released for the first time a dataset of urban traffic objects in two parts: Part A was acquired using an infrared sensing camera in real road environments, and Part B was extracted from Russian news videos. Initially, our primary consideration in searching for a traffic crash dataset was to use a dataset from China exclusively. However, Chinese traffic crash data may contain mosaics in key feature areas, which could significantly hinder TAR-DETR's feature recognition. As a result, we decided to incorporate Russian traffic crash data.

The compressed file contains 3000 original images, along with corresponding labels in both VOC and YOLO formats.