See the MediaEval 2022 webpage for information on how to register and participate.
Task participants create systems that derive insights from multimodal data to understand urban life and air pollution.
The organizers provide two datasets. The first one, namely “environmental data,” contains air quality data such as PM10, PM2.5, CO, NO2, SO2, O3 and weather factors such as temperature, humidity, wind, rainfall, and UV collected from 10 stations spread over Dalat City, Vietnam. The second one, namely “traffic data,” contains videos/images captured from the CCTV system of Dalat city, Vietnam.
The task requires participants to tackle two subtasks:
Participants are encouraged to consult the reference list below.
Improving the quality of human life in smart cities is an important objective linked directly to several United Nations Sustainable Developmental Goals, such as climate action and life on land. Urban transportation and air pollution are two key factors affecting the quality of life. To our knowledge, no literature exists that aims to understand the correlation between traffic factors (e.g., time, vehicles, trees, people), weather (e.g., rain, snow, flood), and air pollution in traffic imagery data (e.g., CCTV, lifelog camera, personal camera). This task aims to encourage participants to develop a novel and generic framework that can discover correlation (or association) between various traffic factors, weather, and air pollution in a locality. By utilizing these correlations, we aim to enhance the accuracy of AQI prediction and the ability to understand the mutual impact between urban life and Air pollution.
This task targets (but is not limited to) researchers in the areas of multimedia information retrieval, machine learning, AI, data science, data mining, knowledge discover, event-based processing and analysis, multimodal multimedia content analysis, lifelog data analysis, urban computing, urban management, environmental science, and atmospheric science.
The task introduces a novel dataset that includes weather and air quality data from ten stations and traffic data from the CCTV system. The datasets are published online at a specific website.
The weather and air quality are recorded every five minutes, including sensor’s profiles (SensorID, SensorCode, SensorName, Latitude, Longtitude, Altitude), recorded time (Date,Time), weatther (Temperature, Humidity, WindSpeed, WindGust, Direction, Rainfall), and air quality (PM1.0, PM2.5, PM10, CO, NO2, SO2, O3, UV). The traffic data contains video streaming from CCTV system. Nevertheless, the archive traffic data contains only one frame per five seconds.
The data is now online, and the participants can access the server to crawl data. Hence, the participants can prepare the ground truth by themselves.
The evaluation of the results submitted by participants will carry on as follows:
For example, July 6 is the date participants have to predict the AQI, and July 1 is the day to submit their predicted AQI values.
For each subtask, the evaluation method is applied as follows:
Here are several research questions related to this challenge that participants can strive to answer in order to go beyond just looking at the evaluation metrics:
Please contact your task organizers with any questions on these points.
 Tuan-Vinh La, Minh-Son Dao, Kazuki Tejima, Rage Uday Kiran, Koji Zettsu: Improving the Awareness of Sustainable Smart Cities by Analyzing Lifelog Images and IoT Air Pollution Data. IEEE BigData 2021: 3589-3594
 Minh-Son Dao, Koji Zettsu, Rage Uday Kiran: IMAGE-2-AQI: Aware of the Surrounding Air Qualification by a Few Images. IEA/AIE (2) 2021: 335-346
 Vo, P.B., Phan, T.D., Dao, M.S., Zettsu, K.: Association Model between Visual Feature and AQI Rank Using Lifelog Data, IEEE Big Data 2019, pp. 4197-4200
 Dat Q. Duong, Quang M. Le, Tan-Loc Nguyen-Tai, Hien D. Nguyen, Minh-Son Dao, Binh T. Nguyen: An Effective AQI Estimation Using Sensor Data and Stacking Mechanism. SoMeT 2021: 405-418
 So Nakamura, R. Uday Kiran, Palla Likhitha, Penugonda Ravikumar, Yutaka Watanobe, Minh-Son Dao, Koji Zettsu, Masashi Toyoda: Efficient Discovery of Partial Periodic-Frequent Patterns in Temporal Databases. DEXA (1) 2021: 221-227
 Ngoc-Thanh Nguyen, Minh-Son Dao, Koji Zettsu: Complex Event Analysis for Traffic Risk Prediction based on 3D-CNN with Multi-sources Urban Sensing Data. IEEE BigData 2019: 1669-1674