Vehicle Tracking Dataset

To support the research, development, and evaluation of vehicle tracking systems, we collected and constructed two sub-datasets: AICC22-Custom.This sub-dataset is built upon a video provided by the organizers of the AI City Challenge 2022 for Challenge Track 2: Tracked-Vehicle Retrieval by Natural Language Descriptions. In contrast to the original task, where vehicle trajectories are associated with natural language descriptions, all cars appearing in the video frames are annotated with frame IDs, track IDs, and bounding boxes. VnTraffic.To reflect real-world traffic conditions in Vietnam, where motorbikes dominate most urban roads, a video was collected from the official YouTube channel of HTV – Hanoi Radio and Television Station. All vehicles, including cars and motorbikes, appearing in the video frames are annotated in the same manner.