Lastest research work published in journal: Applying State-of-the-Art Deep-Learning Methods to Classify Urban Cities of the Developing World

This paper shows the efficacy of a novel urban categorization framework based on deep learning, and a novel categorization method customized for cities in the global south. The proposed categorization method assesses urban space broadly on two dimensions—the states of urbanization and the architectural form of the units observed. This paper shows how the sixteen sub-categories can be used by state-of-the-art deep learning modules (fully convolutional network FCN-8, U-Net, and DeepLabv3+) to categorize formal and informal urban areas in seven urban cities in the developing world—Dhaka, Nairobi, Jakarta, Guangzhou, Mumbai, Cairo, and Lima. Firstly, an expert visually annotated and categorized 50 × 50 km Google Earth images of the cities. Each urban space was divided into four socioeconomic categories: (1) highly informal area; (2) moderately informal area; (3) moderately formal area, and (4) highly formal area. Then, three models mentioned above were used to categorize urban spaces. Image encompassing 70% of the urban space was used to train the models, and the remaining 30% was used for testing and validation of each city. The DeepLabv3+ model can segment the test part with an average accuracy of 90.0% for Dhaka, 91.5% for Nairobi, 94.75% for Jakarta, 82.0% for Guangzhou city, 94.25% for Mumbai, 91.75% for Cairo, and 96.75% for Lima. These results are the best for the DeepLabv3+ model among all. Thus, DeepLabv3+ shows an overall high accuracy level for most of the measuring parameters for all cities, making it highly scalable, readily usable to understand the cities’ current conditions, forecast land use growth, and other computational modeling tasks. Therefore, the proposed categorization method is also suited for real-time socioeconomic comparative analysis among cities, making it an essential tool for the policymakers to plan future sustainable urban spaces.

For more details, please read our paper here.

Data & Design Lab

Welcome to Data and Design lab. Here we blend Data science, machine learning, statistical inference with human-centric design principals to aid policymakers what they do the best – policymaking. We believe evidence-based policymaking can help policy-making easier and efficient. However, we do not think everything can be solved from data-driven knowledge. We, therefore, focus only on certain fields- energy, education, ICT, health- where there is an access to data that can be harnessed. 
 
Data science- encompasses ideas and thoughts from computational social science, machine learning, statistical learning, social science, behavioral science and many other new disciplines of physical and social sciences. As the world is embracing digital technologies, machines and humans are producing more data that may directly or indirectly give insights about our environment. Policymakers be that from public offices or from private enterprises may use these insights to shape their policies or business goals. Harnessing data efficiently thus can open up new horizons. The developed world has embraced the idea and many countries have taken initiatives to find out the best way to harness data for public good. Developing world should not wait as this may be the cause of new digital divide as data begets information and information begets efficient business and policy. 
 
Design thinking helps to make understanding of our surroundings better. The researchers believing in Design centric thinking argue that better design can help get a better result, and design flaws may cause a worse result. We believe working with data needs a marriage with Design thinking as we need to think about the consequence of our works before we delve into data-centric research. We believe design thinking, therefore, can play a vital role in public policy. 
 
With the advent of efficient hardware to store and process a large amount of data, computational mechanisms to learn patterns from large datasets and people’s interest to harness data new possibilities of Data science for social good are engendering. We harness state-of-the-art knowledge from social science, arts, physical science and behavioral science and blend them with our visions of human-centric computing to harness insights from Data. 

Research Focus

Connectivity

Education & ICT

Energy

Society & Economy

Sustainable City

Recent News

Prof. Moinul Zaber article titled “How can we start the open data revolution?” has been published in The Daily Star. You can read it online on their website using this link.

(image retrieved from link)

Prof. Moinul Zaber writes about the importance of indendence to facilitate innovation in telecom marketplace in his latest article for The Daily Star. You can read it online using the link below.

https://www.thedailystar.net/views/opinion/news/btrcs-independence-key-innovative-telecom-marketplace-2234031

(image retrieved from link)

Our paper on “Applying State-of-the-Art Deep-Learning Methods to Classify Urban Cities of the Developing World” has been accepted in MDPI Sensors (IF 3.575), special issue Urban Information Sensing for Sustainable Development. This work is the result of a collaboration between Data & Design Lab, AGenCy Lab (Independent University, Bangladesh) and Center for Spatial Information Science (University of Tokyo).

For more details, please check our paper. 

Prof. Moinul Zaber will be joining as one of the speakers for roundtable session for “Data-driven governance in the social sector: opportunities and challenges in a changing world” track in ICEGOV 2021 that will be held in Athens, Greece from 6-8 October, 2021.

For more details, please visit here. 

(image from ICEGOV2021 website)

Prof. Moinul Zaber’s lecture on “The prospect of Non Traditional Data and Computational Social Science for Sustainable Development” is now available in the WFEO’s website. 

Asif Zaman’s paper “Seasonal, Temporal and Spatial Variation of Particulate Matter Concentration in Bangladesh: A Longitudinal Analysis” has been accepted in the 2021 IEEE Region 10 Symposium (TENSYMP 2021) conference.

Prof. Moinul Zaber talks about digital life and privacy in his latest article titled “Do we care about Digital Privacy” on Man and Machine. You can read it on The Daily Star using the link below.

https://www.thedailystar.net/views/opinion/news/do-we-care-about-digital-privacy-2131796

(image retrieved from link)

Recent Works

A. Zaman, S.B. Rabbani, R.R. Haque, M. Zaber. Seasonal, Temporal and Spatial Variation of Particulate Matter Concentration in Bangladesh: A Longitudinal Analysis. TenSYMP 2021.

This paper shows the efficacy of a novel urban categorization framework based on deep learning, and a novel categorization method customized for cities in the global south. The proposed categorization method assesses urban space broadly on two dimensions—the states of urbanization and the architectural form of the units observed. This paper shows how the sixteen sub-categories can be used by state-of-the-art deep learning modules (fully convolutional network FCN-8, U-Net, and DeepLabv3+) to categorize formal and informal urban areas in seven urban cities in the developing world—Dhaka, Nairobi, Jakarta, Guangzhou, Mumbai, Cairo, and Lima. Firstly, an expert visually annotated and categorized 50 × 50 km Google Earth images of the cities. Each urban space was divided into four socioeconomic categories: (1) highly informal area; (2) moderately informal area; (3) moderately formal area, and (4) highly formal area. Then, three models mentioned above were used to categorize urban spaces. Image encompassing 70% of the urban space was used to train the models, and the remaining 30% was used for testing and validation of each city. The DeepLabv3+ model can segment the test part with an average accuracy of 90.0% for Dhaka, 91.5% for Nairobi, 94.75% for Jakarta, 82.0% for Guangzhou city, 94.25% for Mumbai, 91.75% for Cairo, and 96.75% for Lima. These results are the best for the DeepLabv3+ model among all. Thus, DeepLabv3+ shows an overall high accuracy level for most of the measuring parameters for all cities, making it highly scalable, readily usable to understand the cities’ current conditions, forecast land use growth, and other computational modeling tasks. Therefore, the proposed categorization method is also suited for real-time socioeconomic comparative analysis among cities, making it an essential tool for the policymakers to plan future sustainable urban spaces.

A. Zaman, S.B. Rabbani, R.R. Haque, M. Zaber. Seasonal, Temporal and Spatial Variation of Particulate Matter Concentration in Bangladesh: A Longitudinal Analysis. TenSYMP 2021.

Particulate matters having diameters of 2.5 micrometers or less (PM2.5) have been linked with life threatening health issues worldwide. Data centric approach to ascertain the patterns in the propagation of PM2.5 materials in the atmosphere of a region can help policy makers take informed decisions to take proper action. In this paper, we analyze and identify seasonal, hourly, and regional patterns of PM2.5 propagation in Bangladesh from 2017 to 2020 using the Berkeley Earth dataset. We observe that the concentration of PM2.5 particles has a nationwide median value of about 50 μgm -3 , which is unhealthy for sensitive individuals. The concentration varies seasonally and diurnally. We observe that the concentrations of PM2.5 in the air is around five times more in winter than in summer. The mean PM2.5 concentration inside Dhaka is significantly worse around 70 μgm -3 , which is 1.25 times than the average concentration throughout Bangladesh. We also observe average concentration dropped during the covid-19 pandemic due to lockdown. Using cross correlation analysis, we observed how spikes in PM2.5 concentration levels in one zone may correspond with peaked concentrations in a different zone a few hours later, indicating that air currents may cause the particles to move in certain directions. Our exploratory analysis serves as the first cross-country data centric study of the state and propagation patterns of PM2.5 particles within Bangladesh and our findings can serve as foundation for further research on the topic. 

Sayed, M. A., & Zaber, M. I. (2020). Just-in-Time Educational Aid to Deliver Instant Help for Students in Developing Countries. Recent Trends in Information Technology and its Application, 3(3).

A well-structured inland waterways system should help Bangladesh fulfill SDG goals. In this study, we employ complex network analysis methods to analyze the river-port network of the country. We ascertain different types of ports based on their importance and placement in the connectivity network. Data regarding port location, vessel routes, and schedules were collected from governmental resources. Using the data, a connectivity network was built for further analysis. Different measures of network analysis are used to categorize the ports and the network model has been identified. These categories should help transportation planners and policymakers to better design the inland waterways network of Bangladesh.

Contact Us