Overview

What is Tel2Veh?

Tel2Veh is a new dataset comprising Geographical Cellular Traffic (GCT) flows from 49 locations and camera-detected vehicle flows at 9 of these locations, facilitating the task of leveraging cellular traffic data for vehicle flow prediction. The goal is to utilize the extensive coverage of GCT flows to predict vehicle flows in camera-free locations, thereby addressing the limitations in traffic analysis caused by the restricted scope of device deployment.

Download Data

Dataset

Geographical Cellular Traffic Flow

GCT (Geographical Cellular Traffic) is cellular traffic data associated with its generated location. With its spatial attribute, we can flexibly select regions for data collection. GCT flow refers to the accumulated GCT within a region over a specific time interval (typically 5 minutes). Over time, the series of GCT flows reflect the patterns of the respective region.

GCT Flow Examples

Camera-based Vehicle Flow

We installed six cameras on chosen road segments and utilized BOT-Sort to track and count vehicle flow. Considering Taiwan’s distinctive traffic conditions, such as dense motorcycle traffic, we fine-tuned YOLOv7 and FastReID used in BOT-Sort, with over 1,000 manually labeled instances for robust detection accuracy.

Vehicle Flow Collection

Playground for Dataset Visualization
Please click on the icon to display the data.

Authors

ChungYi Lin

Ph.D. candidate in the Graduate Institute of Networking and Multimedia at National Taiwan University

Senior Researcher in the Department of Internet of Things at Chunghwa Telecom Laboratories.

Shen-Lung Tung

Ph.D. from the Department of Electrical Engineering at National Central University