COMPASSFor the poster presentation, we recommend adhering to the following format:
- Size: The poster should be printed in A0 size (841 mm x 1189 mm). The poster can be printed in portrait or landscape orientation. This standard size ensures optimal visibility and readability for the attendees.
- Legibility: You should ensure that every text in the poster is legible. We recommend that no font be less than size 18 (including font used in labels and axes of various charts, images/diagrams, etc.) should be used.
The template to make your poster is attached here.
| Poster Title | Authors | Affiliations | Abstract |
|---|---|---|---|
| A Deep Learning Approach to Detect Severity of Mango Damage in the Early Fruit Stage | Yonasi Safari, Joyce Nakatumba-Nabende, Rose Nakasi, Rose Nakibuule, Simon Allan Achuka | Mbarara University of Science and Technology, Makerere University, Makerere University, Makerere University, Makerere University | Accurate detection of mango fruit damage predominantly from fruit fly infestation is pivotal as it directly affects both yield and trade worldwide. Therefore, timely identification of such damage is critical to mitigating the spread of infestation and minimizing as- sociated losses. This paper focuses on the early detection of mango damage in orchards using YOLOv8 models, that offer enhanced accuracy and speed compared to earlier versions, making them more efficient for object detection tasks. Limited studies have been done to detect and classify damage on fruits in orchards using deep learning, with the need for more models to detect various categories of damage instances. The experiments in this study revealed no sub- stantial differences among the various YOLOv8 versions used with the highest accuracy of 88.6% and 98.5% attained for detecting dam- age and mango instances respectively. Both YOLOv8s and YOLOv8l obtained a precision value of 88.6% for lesion detection, and 87.9% using YOLOv8x. However, YOLOv8x achieved slightly higher values of recall and mAP compared to other models in detecting damage features. The study has further revealed that learning the damaged features of mango fruit is more challenging compared to healthy features, as observed from values obtained from the precision-recall curve. Through fine-tuning parameters of the models, our experi- mental results using the YOLOv8 model demonstrate the potential of lesion detection on mango fruits on trees, leveraging a dataset of 1317 images augmented to 3161. This study addresses the challenge of estimating profits and losses for fruits still on trees, which has been relatively overlooked in prior research efforts. We believe this method can effectively be adapted to detecting lesions on other fruits in orchards with minimal modifications. Future work can con- sider a better dataset with minimal noise while exploring different growth stages of fruits, and weather conditions the data is captured using alternative models while incorporating other factors in the segmentation and analysis phases. |
| A Hierarchical Framework For Tree Classification Using Satellite Time Series Data | Chilukamari Shiva Sai Krishna, Dhruvi Goyal, Aaditeshwar Seth | Indian Institute of Technology Delhi, Indian Institute of Technology Delhi, Indian Institute of Technology Delhi | Biodiversity of tree plantations plays an important role in soil conservation and for forest health. It not only improves soil and forest health, but also increases the carbon captured significantly in the form of soil biomass. There is a need to assess the biodiversity or the types of tree species present in such cases. The current work proposes a framework to perform hierarchical classification of tree species using time series of open-source satellite data. A detailed study is being conducted at IIT Delhi campus, to build a framework for hierarchical tree classification, that demonstrates the usability and scope of this approach. The proposed mechanism is robust and can be extended to any region, for classifying the trees. |
| Archives, Environmental Knowledge, and Epistemic Injustices in the Himalayas | Aarjav Chauhan | University of Toronto | The Himalayan region serves multiple functions for the commu- nities residing within them, including harboring biodiversity, sup- porting diverse cultures and livelihoods, and providing natural resources. The environmental and social impacts of climate change in this vulnerable region create uncertainty for the livelihoods and indigenous practices of local communities. We argue that this un- certainty should not be perceived negatively, but rather viewed from a perspective that embraces multi-faceted forms of knowl- edge production and practice. Our position is that understanding the impacts of climate change in the Himalayas starts with un- derstanding how knowledge is produced and legitimized. Digital archives, organized and managed collections of digital information, are information infrastructures that present a mode of convergence between the social and material bases of knowledge work and the relations of the participating people who produce knowledge. An effective digital archive serves the needs, activities, and contexts of the people who use, create, and contribute to it. Our ongoing research evaluates existing digital archives of the Himalayan re- gion to understand the social, technical, and political boundaries within which these archives function. Drawing on content analysis and interviews with archive creators, curators, and participating communities, we question how studying digital archives within culturally diverse regions, such as the Himalayas, can inform our understanding how local environment knowledge is legitimized. |
| Capturing Human Emotion Pervasively using COTS mmWave Radar | Argha Sen, Amrta Chaurasia, Avijit Mondal, Sandip Chakraborty | Indian Institute of Technology Kharagpur, Google Bangalore, Intuit Bangalore, Indian Institute of Technology Kharagpur | Emotion recognition has emerged as a crucial field with diverse applications in medical, educational, entertainment, marketing, and advertising domains. Detecting emotions can facilitate early inter- vention in medical care, enable real-time adjustments in interactive gaming, and personalize smart home environments. Traditional methods for emotion recognition primarily rely on audio-visual cues or physiological signals, which can be intrusive, inconvenient, or limited in their ability to capture subtle emotional changes. To address these challenges, we propose Millimeter Wave (mmWave)- based non-contact, passive sensing for human emotions. The pro- posed method uses a Convolutional Neural Network (CNN) to predict four different emotion classes. Our results demonstrate the potential of mmWave technology for developing accurate and ef- ficient human emotion recognition systems in various real-world applications. |
| Determinants of lease rents of village ponds for aquaculture: Evidence from central Gujarat | Arnab Paul Choudhury, Philip Kuriachen, Shilp Verma, Subhodeep Basu, Tushaar Shah | International Water Management Institute Anand, International Water Management Institute Anand, International Water Management Institute Anand, International Water Management Institute Anand, International Water Management Institute Anand | The fishery sector in the country is undergoing a remarkable trans- formation with India now currently being the second largest fish producer in the world along with fish becoming the highest ex- ported item from India. Despite the sector’s growth, challenges per- sist for smallholders that limit their productivity ultimately curbing livelihoods opportunities. Maximising profitability is imperative amidst rising input costs. One key factor the study identified in this study is the lease rent paid for community ponds allocated to farm- ers. By focussing on Central Gujarat’s Village Pond leasing policy the study examines the dynamics surrounding the auctioning of vil- lage ponds and the factors influencing the same. Contrary to other states, Gujarat employs an open auction policy for village ponds with the objective of encouraging competitive bidding among inter- ested parties. But because of prevalence of asymmetry information on pond values among bidders, the efficacy of auctions gets hin- dered leading to inefficient allocation. The study hypothesizes that auctions for village ponds resemble common value auctions, where bidders face uncertainty regarding the true value of the asset. Iden- tifying the determinants of bidder valuation and integrating them into the determination of upset prices can mitigate information asymmetry, enhance auction efficacy, and reduce adverse selection. The study also identified considerable variance between the bid values of different ponds. Bid value on average as per the study comprises 25 % of the total cost for a fish farmer in Anand district of Gujarat and there is tremendous difference between the highest bid value/hectare and the lowest bid value/hectare. While the highest bid value of the surveyed pond was Rs 1.7 lakh/hectare, the lowest bid value was recorded at Rs 900/hectare. This huge variance in the bid value data and factors contributing was considered a cru- cial area of exploration. The study has deployed Directed Acyclic Graphs and Linear Regression as methodologies to identify the factors primarily influencing the bid values and to measure the extent of that influence. The key findings could inform strategies to address such cost constraints, ultimately empowering smallholders to capitalize on opportunities within this rapidly evolving sector. By optimizing the auction process, stakeholders can unlock the full potential of village ponds, ultimately driving inclusive growth and sustainable development in India’s fisheries sector. |
| Developing a Communication Application for Children with Autism through a Participatory Design Approach | Mayisha Muntaha Alam, Shams Akbar Aalok, Farhan Ahmed Fahim, Sami Uddin, Obayed Ur Rahman, Marzan Mahatab, Ashfaq Mahee Siddiky, Md. Tamjid Hosain, Anik Sinha, Nova Ahmed | North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka, North South University Dhaka | In developing regions, the lack of cultural context-based technological intervention with means of communication between parents and teachers with Children with Autism (CWA) led us to design and develop an application to address this underexplored area. Three participants (two parents and one teacher of CWA) participated in two participatory design sessions to develop and improve the application. The same three participants shared their views on the design and features of the application in the first study and provided feedback about them in the second study. The study aims to bridge the gap with a communication medium between the CWA and their parents and teachers with an application that can be customized and personalized to cater to the user's needs. This work aligns with the contextual design factor of HCI researchers by designing a communication platform for CWA. |
| Exploiting Air Quality Monitors to Perform Indoor Surveillance: Academic Setting | Prasenjit Karmakar, Swadhin Pradhan, Sandip Chakraborty | Indian Institute of Technology Kharagpur, Cisco Systems, Indian Institute of Technology Kharagpur | Due to public awareness and government regulations, low- cost air quality monitors are becoming ubiquitous in modern indoor spaces. These monitors primarily sense air pollutants to augment the end user’s understanding of her indoors. Studies have shown that having access to one’s air quality context reinforces the user’s need to take necessary actions to improve the air over time. Hence, user’s activities signifi- cantly influence the indoor air quality. Such correlation can be exploited to get hold of sensitive indoor activities from the side channel air quality fluctuations. In this study, we explore the odds of identifying eight different indoor activities (i.e., enter, exit, fan on, fan off, AC on, AC off, gathering, eating) in a research lab with the help of an in-house low-cost air quality monitoring platform named DALTON. Our extensive data collection and analysis over the three months shows 97.7% overall accuracy in our dataset. |
| Exploring the role of Generative AI in supporting Women’s Aspirations for Learning Programming during Socio-Political Conflict in Afghanistan | Vikram Kamath Cannanure, Hamayoon Behmanush, Freshta Akhtary, Roghieh Nooripour, Ingmar Weber | Saarland Informatic Campus - Saarland Unviersity, Saarland Informatic Campus - Saarland Unviersity, Parwan University, Alzahra University, Saarland Informatic Campus - Saarland Unviersity | Due to the socio-political conflict in Afghanistan, stringent bans on education and employment for women have deeply impacted their aspirations. This study explores how online education and Generative AI play a role in these challenges by surveying 136 fe- male programming students in Afghanistan. We assess participants’ aspirations and engagement with Generative AI in the context of learning programming. Our findings reveal that participants are changing their careers to online software development and remote work to adapt to the political change. Moreover, participants use Generative AI as a support tool to study programming in the ab- sence of resources. Lastly, we adapted a survey scale based on the hope-scale for aspirations to learn programming. Our work opens design opportunities for AI to enhance the learning programming experience amid socio-political conflict. |
| Eye in the Sky: Detection and Compliance Monitoring of Brick Kilns using Satellite Imagery | Rishabh Mondal, Shataxi Dubey, Vannsh Jani, Shrimay Shah, Suraj Jaiswal, Zeel B Patel, Nipun Batra | Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar | Air pollution kills 7 million people annually. The brick manufac- turing industry accounts for 8%-14% of air pollution in the densely populated Indo-Gangetic plain. Due to the unorganized nature of brick kilns, policy violation detection, such as proximity to human habitats, remains challenging. While previous studies have utilized computer vision-based machine learning methods for brick kiln de- tection from satellite imagery, they utilize proprietary satellite data and rarely focus on compliance with government policies. In this research, we introduce a scalable framework for brick kiln detection and automatic compliance monitoring. We use Google Maps Static API to download the satellite imagery followed by the YOLOv8 model for detection. We identified and hand-verified 19579 new brick kilns across 9 states within the Indo-Gangetic plain. Furthermore, we automate and test the compliance to the policies affecting human habitats, rivers and hospitals. Our results show that a substantial number of brick kilns do not meet the compliance requirements. Our framework offers a valuable tool for govern- ments worldwide to automate and enforce policy regulations for brick kilns, addressing critical environmental and public health concerns. |
| HumekaFL: Automated Detection of Neonatal Asphyxia Using Federated Learning | Pamely Zantou, Blessed Guda, Bereket Retta, Gladys Inabeza, Carlee Joe-Wong, Assane Gueye | Carnegie Mellon University Kigali, Carnegie Mellon University Kigali, Carnegie Mellon University Kigali, Carnegie Mellon University Kigali, Carnegie Mellon University Pittsburgh, Carnegie Mellon University Kigali | Birth Apshyxia (BA) is a severe condition characterized by insuf- ficient supply of oxygen to a newborn during the delivery. BA is one of the primary causes of neonatal death in the world. Al- though there has been a decline in neonatal deaths over the past two decades, the developing world, particularly sub-Saharan Africa, continues to experience the highest under-five (<5) mortality rates. While evidence-based methods are commonly used to detect BA in African healthcare settings, they can be subject to physician errors or delays in diagnosis, preventing timely interventions. Centralized Machine Learning (ML) methods demonstrated good performance in early detection of BA but require sensitive health data to leave their premises before training, which does not guarantee privacy and security. Healthcare institutions are therefore reluctant to adopt such solutions in Africa. To address this challenge, we suggest a federated learning (FL)-based software architecture, a distributed learning method that prioritizes privacy and security by design. We have developed a user-friendly and cost-effective mobile appli- cation embedding the FL pipeline for early detection of BA. Our Federated SVM model outperformed centralized SVM pipelines and Neural Networks (NN)-based methods in the existing literature. |
| Improving Performance of Fine-Grained Air Quality Modeling with Domain-Inspired Predictors: A Case Study over Delhi, India | Zeel B Patel, Nipun Batra, Sarath Guttikunda | Indian Institute Of Technology Gandhinagar, Indian Institute Of Technology Gandhinagar, UrbanEmissions.info | Air pollution kills 7 million people every year. Na- 2 tions have deployed air quality sensors but they are 3 costly and thus have spatially sparse coverage. Air 4 quality modeling (interpolation, forecasting) stud- 5 ies have been conducted in the past to increase 6 spatial coverage, but i) physics-driven models do 7 not take sensor readings into account and thus ex- 8 hibit high bias and ii) machine learning models fo- 9 cus mainly on empirical performance and lack ac- 10 tionable insights, such as, potential reduction in 11 air pollution by removing a source. In this pa- 12 per, we present an experimental case study on a 13 fine-grained curated dataset for machine learning 14 based PM2.5 modeling. We describe the proxies 15 we combine with sensor data for various sources 16 of air pollution including but not limited to popu- 17 lation, fire counts and road networks. We show the 18 improvement in modeling performance by includ- 19 ing these additional domain-inspired variables; and 20 how these allow us to study what-if scenarios to- 21 wards mitigation. |
| Inducing Socially Optimal Vaccination Behaviour through Vaccination Subsidies in Epidemics with Reinfection | Amal Roy, Chandramani Singh, Y Narahari | Indian Institute of Science Bengaluru, Indian Institute of Science Bengaluru, Indian Institute of Science Bengaluru | In this paper, we investigate the following research question: In a population consisting of rational and intelligent individuals who make their own individual decisions regarding vaccination (to vac- cinate or not), how can a policy maker incentivize individuals to achieve socially optimal vaccination behaviour? For this, we work with an extended SIR (Susceptible-Infected-Recovered) model of epidemics in the presence of vaccinations and with possibility of reinfections captured explicitly. We study two models: (1) a cen- tralised, optimal control model yielding a socially optimal vaccina- tion policy and (2) a decentralised game theoretic model yielding an individually optimal vaccination policy. To achieve feasibility and computational tractability of solving these models, we use a mean field approach, leading to a mean field control (MFC) model and a mean field game (MFG) model, respectively. Our study shows that socially optimal behavior can be achieved in a rational population through appropriate vaccination subsidies. This will be a valuable input for the policy maker to decide on the quantum of subsidy to be provided to induce social optimal vaccination behaviour in a rational population. |
| Lessons from Skill Development Programs - Livelihood College of Dhamtari, Chhattisgarh, India | Arnab Paul Choudhury, Nihal Patel | Viksit Labs Foundation, Indian Institute of Technology Guwahati | India has one of the youngest workforce globally which puts it at an advantage on the demographic front where a majority of its inhabi- tants are able-bodied workers. This demographic dividend has been touted to fuel economic growth and has the potential to improve the standard of living in the country. However, without proper educa- tion and training of these large masses of able-bodied workers, this demographic dividend would largely remain unutilized hence the Govt of India initiated a multitude of Skill Development programs with a desire to enable its inhabitants so that they may secure a better work-life balance, job security, career growth opportunities, and a healthy work environment. Schemes such as the “Skill India Initiative” of the Central Government of India and the “Livelihood College” model adopted by the state of Chhattisgarh are aimed at making strides in this direction. In particular, Livelihood Colleges are aimed at training and enabling a secured livelihood for blue- collar jobs which are often less adaptable to the dynamic economic changes in the market. In this particular study, we look into the 5-stage skill development process adopted in a Livelihood College located in the Dhamtari district of Chhattisgarh to understand and further improve upon the service delivery of such programs that are implemented throughout the state of Chhattisgarh as well as across India. We engaged in informal discussions with various stakehold- ers involved in Dhamtari’s Livelihood College for over a year to understand the capacities and constraints throughout the skill de- velopment process. The prolonged exposure allowed us to identify two major challenges. First, a lack of inclusive mobilization and counseling process, and second, inconsistent student attendance or presence in classrooms. Following these qualitative inputs we conducted a quantitative analysis of students and an analysis of CCTV streams to further verify these challenges. We also briefly discuss the concerns and limitations of our current study and call for further work in this direction. |
| Live In-car Traffic Monitoring using mmWave Sensing | Rajib Sarkar, Argha Sen, Sandip Chakraborty | Indian Institute of Technology Kharagpur, Indian Institute of Technology Kharagpur, Indian Institute of Technology Kharagpur | Traffic monitoring is crucial for managing urban transportation sys- tems and ensuring road safety. Traditional methods, such as cameras and inductive loop sensors, have limitations in adverse weather conditions or high traffic densities. In this poster, we propose a novel approach for live, in-car traffic monitoring using Millime- ter Wave (mmWave) radar sensing. Our method utilizes mmWave range-doppler and range-azimuth heatmaps as input for a 2D con- volutional neural network (2D-CNN) to perform live traffic moni- toring. The classification model determines the presence of traffic and classifies the number of vehicles as well as the type of vehicle (two-wheeler, four-wheeler, etc.). We evaluate our approach using real-world mmWave radar data and demonstrate its effectiveness in accurately monitoring traffic conditions in various scenarios. |
| Measuring Public Value in Co-Produced Public Services: The case of separate waste collection in St. Petersburg, Russia | Daria Bakalets | St. Petersburg State University | The paper examines the problem of creating public value for services that involve co-production – cooperation between government organizations and citizens. Expanding on a foundational work by M. Moore (1995), we introduce a modification to the public value creation model to incorporate co- produced public services. Furtherly we applied this modification using the example of a separate waste collection service (SWC) – a service whose effective implementation critically depends on the tight partnership between government authorities and the population, ultimately aiming to improve environmental conditions and foster sustainable development of the area. Using the collected information on 671 respondents, the developed model was estimated using the partial least squares structural equation modeling approach (PLS-SEM). We demonstrate that there is a set of factors that, in their interconnection, statistically significantly increase the public value of SWC. These results shed light on the most effective measures to enhance public participation in waste separation and thus help to prioritize government spending in this domain. |
| On the Promises and Perils of CogSOCs | Shashank Yadav | Indian Institute of Technology Bombay | There is growing trend among cyber Security Operations Centers (SOCs) to also integrate online influence, social botnets and AI-disinformation oriented capabilities in their offerings. This is leading to the creation of specialized components within SOCs, referred to as CogSOCs (Cognitive Security Operations Centers), which are evolving with their own Cyber Threat Intelligence (CTI) standards and sometimes a sociopolitical orchestration logic. In light of the competition to standardise CogSOCs’ CTI and practices, we explore various data policy challenges emerging at syntactic, semantic, and operatic dimensions of CTI orchestration and standardisation in CogSOCs. The article aims to underline the data governance issues in deploying automation to address the increasingly social, and sometimes political, automation in our digital societies and threat environments. |
| Poster: Soil Powered Computing | Harshitha Kakani, Adam Dunlop, Colleen Josephson, Stephen Taylor, John Madden | University of California Santa Cruz, Drexel University, University of California Santa Cruz, University of California Santa Cruz, University of California Santa Cruz | Soil Microbial Fuel Cells (SMFCs) are emerging as a sustainable and eco-friendly technology, harnessing microbial processes to gener- ate electricity in environments where conventional energy sources such as batteries and solar panels are less effective. We present an overview of the opportunities and challenges in the SMFC space, as well as preliminary results on a new technique that has promise for making SMFCs a practical energy source for outdoor sensor net- works.Through systematic deployments and rigorous experiments, we refined SMFC designs to enhance their robustness and energy generation capabilities. The findings demonstrate significant im- provements in line with up to a 40% increase in digital computing operations and a 120% increase in analog sensor runtime compared to traditional configurations. These advancements are substan- tiated by data-driven validation from real-world field deployments, establishing a solid foundation for the practical application of soil- powered computing and promoting the development of sustainable, battery-free sensor networks. |
| South Asian AI Ethics Framework: What Values Are We Looking For? | Farzana Islam, Tasmiah Tahsin Mayeesha, Shahbaz Ahmed, Nova Ahmed | North South University Dhaka, North South University Dhaka, North South University Dhaka | In recent times, the term AI ethics has caught the attention of academics, legislators, developers, and AI users to promote ethical AI development. Based on 36 qualitative interviews with different stakeholders, including machine learning practitioners, academic researchers, and policymakers in the emerging AI ecosystem, we investigated the creation of an AI ethics framework for South Asia. This paper explores the question of what values should underpin such a framework. We examine the ethical considerations specific to the South Asian context, drawing on existing frameworks and considering the region's unique social, cultural, and economic landscape. The paper investigates core values such as fairness, transparency, accountability, and human oversight, analyzing their importance in ensuring responsible AI development and deployment in South Asia. By identifying these crucial values, the paper aims to contribute to the creation of a comprehensive South Asian AI ethics framework that fosters trust, promotes inclusivity, and mitigates potential risks associated with AI technologies. |
| Towards Identifying the Distribution of Eviction from User Review Data on Online Platform | Shruti Nitturkar, Maryam Tabar | University of Texas at San Antonio, University of Texas at San Antonio | The eviction of tenants is an urgent societal challenge in the United States, which could have significant adverse impacts on the individ- ual’s health and well-being. Access to granular information on the distribution of eviction plays a key role in the effective and efficient implementation of eviction mitigation programs, and accordingly, this project focuses on understanding the distribution of evictions from Airbnb review data in New York City (NYC). In particular, we conduct aspect-based sentiment analysis on reviews and compute correlations between the number of evictions and the average sen- timent analysis scores for various aspects in reviews. The result of our preliminary research suggests that the sentiment with respect to each aspect alone may not have strong direct signals regarding the eviction distribution. |
| Towards Sustainable Personalized On-Device Human Activity Recognition with TinyML and Cloud-Enabled Auto Deployment | Bidyut Saha, Riya Samanta, Soumya K. Ghosh, Ram Babu Roy | Indian Institute of Technology Kharagpur, Indian Institute of Technology Kharagpur, Indian Institute of Technology Kharagpur, Indian Institute of Technology Kharagpur | Human activity recognition (HAR) holds immense potential for transforming health and fitness monitoring, yet challenges persist in achieving personalized outcomes and sustainability for on-device continuous inferences. This work introduces a wrist-worn smart band designed to address these challenges through a novel combi- nation of on-device TinyML-driven computing and cloud-enabled auto-deployment. Leveraging inertial measurement unit (IMU) sen- sors and a customized 1D Convolutional Neural Network (CNN) for personalized HAR, users can tailor activity classes to their unique movement styles with minimal calibration. By utilising TinyML for local computations, the smart band reduces the necessity for constant data transmission and radio communication, which in turn lowers power consumption and reduces carbon footprint. This method also enhances the privacy and security of user data by lim- iting its transmission. Through transfer learning and fine-tuning on user-specific data, the system achieves a 37% increase in accuracy over generalized models in personalized settings. Evaluation using three benchmark datasets — WISDM, PAMAP2, and the BandX — demonstrates its effectiveness across various activity domains. Additionally, this work presents a cloud-supported framework for the automatic deployment of TinyML models to remote wearables, enabling seamless customization and on-device inference, even with limited target data. By combining personalized HAR with sus- tainable strategies for on-device continuous inferences, this system represents a promising step towards fostering healthier and more sustainable societies worldwide. |
| Understanding User Preferences for More Sustainable Data Centers from a Digital Sufficiency Lens | Harshit Gujral, Dushani Perera, Christina Bremer, Steve Easterbrook | University of Toronto, Cardiff University, Lancaster University, University of Toronto | Despite energy efficiency improvements, the carbon footprint of data centers is continuously increasing due to a rapid growth in data. Since 26% of their energy consumption is due to cloud storage and servers, which are directly interacted with and managed by users, we argue that effectively reducing their carbon footprint requires designing avenues for user involvement. To better understand users’ storage needs and gauge their willingness to adopt sustainable stor- age practices, we conducted an online survey (N=125). Our findings highlight users’ limited awareness of data centers’ carbon footprint but also their willingness to switch to a more sustainable provider once being made aware. We propose interface design recommen- dations for cloud storage providers that challenge uniform storage policies to address users’ diverse needs and usage patterns. We also argue for integrating sufficiency with interface design to empower users to deliberately manage their data, minimize unnecessary stor- age, and optimize data utilization. |
| Use of Satellite Data to Estimate Small-Scale Tree Plot Canopy Height Distribution and Density in India | Dhruvi Goyal, Harsh Singh Chauhan, Aaditeshwar Seth | Indian Institute of Technology Delhi, Indian Institute of Technology Delhi, Indian Institute of Technology Delhi | This article explores the usage of satellite data from Sentinel 1 and Sentinel 2 to estimate small scale tree plot canopy height distri- bution and canopy cover density in India using machine learning methods. It discusses the need for such a study and talks about the challenges in getting the right ground truth for canopy height and canopy cover density modelling. In the absence of voluminous ground truth, how remotely sensed GEDI data can be used for training the model has been discussed. The methodology has been described clearly and the outputs that will be produced have been mentioned along with a few example output layers. The results of the models of one of the agroclimatic zone have been mentioned. The utility of open-source tools built using the aforementioned methods for communities, non-profit organisations and govern- ment bodies, has also been discussed. |
| Using Satellite Images to Track Relative Socioeconomic Development in India | Aatif Nisar Dar, Badrinath Padmanabhan, Mrunal Atul Kadane, Aaditeshwar Seth | Indian Institute of Technology Delhi, Indian Institute of Technology Delhi, Indian Institute of Technology Delhi | Studying how regions develop over time can offer valuable insights, but it’s challenging with traditional data like censuses and sur- veys, which aren’t frequent and face a lot of delays. Using satellite imagery for socio-economic indicators has become a useful alter- native to track development at fine spatial and temporal scales. In this paper, we train a model using satellite imagery to estimate socio-economic development at the village level in India. We test its consistency over time and use it to analyze development trends over a two-decade period. Our study looks at how factors like the geographic distance of a village to economic hubs and the inequal- ity of development in the district affect village development. Our results provide evidence of the possible impact that policy changes during this period have had on village development. |
| VayuBuddy: LLM-powered natural language interface for exploring and understanding air pollution data | Yash Bachwana, Khush Shah, Nitish Sharma, Zeel B Patel, Nipun Batra, Sarath Guttikunda | Indian Institute of Technology Gandhinagar, Indian Institute of Technology Gandhinagar, None, Indian Institute of Technology Gandhinagar, Indian Institute of Technology Gandhinagar, UrbanEmmissions.info | Almost 6.7 million lives are lost to air pollution each year, making it one of the biggest threats to environmental health. However, the difficulties in communicating about air pollution obstruct public awareness and action. The gaps in the current ecological health literacy further highlight this problem. In this work, we present VayuBuddy: a LLM-powered natural language interface to explore and understand air pollution data. VayuBuddy is a web application that can comprehend user requests and offer customised replies based on the data from the pollution board in India. We create a dataset of natural language prompts and find that LLMs are highly accurate in generating code that gets executed to answer the query from the dataset. Systems can such as VayuBuddy can lower the entry-barrier into understanding and (potentially mitigating) air pollution. |
| Visual Feedback Interface for Audio Communication Over Lossy and High Delay Networks | Heinsamding thou, Bhaskaran Raman, Manish Kumar, Ajith Pasuvula | Indian Institute of Technology Bombay, Indian Institute of Technology Bombay, Indian Institute of Technology Bombay | In an audio communication via VoIP (Voice over Internet Protocol) over a network with high delay and/or loss, participants can step-over each other and have a poor user experience. This work presents a novel approach to mitigate step-overs during audio call and hence enhance user experience in audio communications, particularly in scenarios characterized by high network delays and/or loss. The proposed solution involves the integration of visual feedback into the user interface (UI). Through the use of dynamic animations displayed on the user’s screen, real-time updates regarding the reception status of the transmitted audio, at the intended recipient are provided. This innovative method aims to enhance the user experience during audio communication sessions under adverse network conditions. We employed a combination of qualitative and quantitative methods to validate our hypothesis, which aimed to reduce step-overs during audio calls with high delay, ultimately enhancing the user experience. The results were positive, with participants reporting improved experiences when animation was enabled where the average rating of the Mean Opinion Score (MOS), which ranges from 1 to 5 (with 1 indicating the lowest and 5 indicating the highest satisfaction), increased from 2.6 to 3.7. And we also observed a significant reduction of approximately 35% in step-overs when the animation was enabled ("animation on") as compared to when it was disabled ("animation off"). |