About
With the rapid revolution and increasing availability of geospatial data, not only academia but also industry aspire for solutions to further leverage the big data and AI technologies to create new products, improve efficiencies and provide novel solutions to existing problems. However, despite the widespread interest, there is a lack of communication between the researchers in academia and industry, limiting advancements at the intersection. Academia often has limited access to the rich and potentially useful big geospatial datasets and related real problems. In addition, the solutions proposed by the academic researchers alone are usually developed for small scale with many assumptions, leaving a less-attended gap between methods and their applicability at scale for industrial applications. On the other hand, industry has the data and problems at scale. However, since existing research is often not on par, industry researchers may lean towards using the traditional approaches that are developed without spatial consideration (e.g., ignoring spatial and temporal dependencies), and project teams have limited time and efforts to dive deep on the development of novel techniques that can be high-risk but high-potential. This opens up opportunities for synergistic collaboration between industrial practitioners and academic researchers. The 2nd ACM SIGSPATIAL International Workshop on Spatial Big Data and AI for Industrial Applications (GeoIndustry 2023) is to offer a forum to exchange thoughts and ideas between industry and academia and reduce the siloed efforts. At the same time, the collaborations, via invited and regular talks, can not only accelerate the research-to-impact cycle, but also foster workforce development for future geospatial researchers.
Organization Committee
Program Chairs
| Heba Aly (Amazon) |
| Emre Eftelioglu (Amazon) |
| Song Gao (University of Wisconsin-Madison) |
| Yan Li (Amazon) |
| Jinmeng Rao (Mineral Earth Sciences) |
| Yiqun Xie (University of Maryland) |
Program Committee
| Jiahao Fan (University of Wisconsin-Madison) | Mingquan Chen (Google) |
| Hongxu Ma (Mineral Earth Sciences) | Chenxi Lin (PAII Inc) |
Schedule
Monday, November 13, 2023 - Hamburg, Germany
| Central European Time | Title |
|---|---|
| 13:50 - 14:00 | Opening Remarks |
| 14:00 - 14:15 | Spatial experiment identification (SPEX-ID): a method to identify experimental conditions from spatial information in digital agricultural data and beyond Taylor Aune, Huan Gu, Dagmawi Woldesenbet, Tuyen Le and Jeffery Sauer Abstract: On-farm experiments (OFE) are the cornerstone of evaluating interventions on important outcomes like crop yield, disease resistance, soil fertility, and more. However, prospectively planning and implementing all OFE of interest is challenging in resource-constrained settings. In addition, the quality of an OFE is determined by the spatial arrangement of the treatment conditions. Experimental conditions can exist in digital agriculture data, although there may be no information indicating that an experiment took place. We introduce a novel method that can identify potential experimental arrangements on a field using only the spatial information on the experimental conditions of interest. We call this method spatial experiment identification (SPEX-ID). We explain the method in detail and highlight its ability to identify several common types of spatial arrangements in on-farm experiments. We discuss the potential of this method in a large sample of nearly 90,000 fields from a large commercial digital agriculture database where the intervention of interest was the application of fungicide. From this sample we were able to identify more than 12,000 fields with potential experimental conditions. None of these fields were previously known to contain experimental conditions. We highlight several examples of subfield regions with high-quality experimental arrangements and discuss several avenues for future research. |
| 14:15 - 14:30 | Floorplan Generation from Noisy Point Clouds Anselmo Talotta, Valentin Radu and Lorenzo Sorgi Abstract: Floorplans are useful for navigating indoor spaces, for resource allocation and management of indoor spaces. But in the absence of readily available floorplan drawings, these are hard to generate. In this work, we enhance a solution for generating floorplans from orderly point clouds to be more robust to noisy measurements. With this approach, point clouds are converted into a density map, which is then used for detecting the shape of rooms and relevant landmarks, such as doors and windows. We improve the robustness of the room detector by training in two stages, firstly using synthetic data with point clouds extracted from 3D graphical representations of indoor spaces; and secondly, extending the training of the model in a new domain by using real-world data collected with Tango devices and their vision based depth estimation. This training gets the domain closer to our final goal, of generating floorplans from easily collected point cloud scans in the real-world. Finally, we showcase the capability of our solution when operating with noisy Lidar scans collected from a drone with posture estimation. |
| 14:30 - 14:45 | A new approach to assessing perceived walkability: Combining Street View Imagery with Multimodal Contrastive Learning Model Xinyi Liu, James Haworth and Meihui Wang Abstract: Walkability is becoming increasingly important in urban planning, public health, and environmental protection. Traditional assessment tools like streetscape images and semantic segmentation focus on objective factors, while questionnaires as the main tool for perceived walkability are limited by cost and scale. This study introduces a new method using the Multimodal Contrastive Learning Model, CLIP, to assess perceived walkability by analysing both tangible and subjective factors such as safety and attractiveness. The method compares perceived with physical walkability by scoring street view images with a customized scale. Initial results indicate CLIP can identify pedestrian-friendly streetscapes that might score low on physical metrics. While its accuracy needs more evaluation, CLIP offers a cost-effective alternative without needing extensive labelled datasets. This method can be combined with objective pedestrian assessment methods to serve as reference information for various industries such as real estate, transportation planning, and tourism. |
| 14:45 - 15:00 | Exploring The Use of OpenStreetMap Data (OSM) and GPS Traces for Validating Driving Routes and Identifying Prohibited Maneuvers in Direction Services Antonios Karatzoglou, Tijana Bekic, Mohit Khanna, Vashutosh Agrawal, Aleksandar Matejevic, Varun Kakkar, Nikola Perin, Jacob Whitbeck, Dragomir Yankov, Michael R. Evans, Nikola Todic, Goran Predovic and Aleksandar Samardzija Abstract: As people increasingly adopt a more flexible and mobile way of living, map applications and navigation services have gained increasingly in importance. In order for these services to be useful to their users, their providers need to ensure and maintain a high quality in terms of accuracy and reliability. Keeping track of the service quality is essential, especially in a dynamic domain like this, where the underlying road network data might change from one day to the next due for instance to temporal road closures or turn restrictions. Although the changes might be small and local, they can still have a big impact on the routing quality of direction services. The focus of this study lies on providing the means to avoid such high-impact scenarios in a cost-efficient manner. In particular, we focus on the legality aspect of routes that reflects the degree of prohibited by law maneuvers within a suggested route. First, we define a set of appropriate sample-based metrics to help us track a route's legality. Then, we introduce an automated pipeline based on OSM data and GPS traces that is able to support and reduce the load of our human judges to identify illegal maneuvers in a sample of candidate routes. Finally, we evaluate our pipeline using a random sample of 1,306 US routes while a local map data provider as well as human judgement via visual inspection serve as our ground truth. Our results show that our method is able to sufficiently cover most of the route sections in our candidate route samples and identify 90% of all illegal maneuvers while reducing the load of manual human judgement by 86%. |
| 15:00 - 15:30 | Coffee Break |
| 15:30 - 15:45 | LocFree: WiFi RTT-based Device-Free Indoor Localization System Mohamed Mohsen, Hamada Rizk, Hirozumi Yamaguchi and Moustafa Youssef Abstract: Gaining knowledge of a person's movement in the environment without needing a specialized device raises the need for device-free indoor localization systems that could be leveraged in many applications including IoT, security, etc. Wi-Fi is one of the most widely adopted technologies for indoor location determination tasks due to its ubiquity. Existing device-free indoor localization systems rely on the utilization of the Received Signal Strength Indicator (RSSI) and the Wi-Fi Channel State Information (CSI). However, RSSI is highly sensitive to environmental noise such as multi-path interference and fading which causes degradation in the system's performance. In addition, CSI suffers a lack of standardization which necessitates the requirement for special hardware or software. In this work, we present LocFree, a deep-learning-based device-free indoor localization system that handles the challenges of RSSI and CSI by leveraging the Time of Flight (ToF) information obtained using the IEEE 802.11mc Fine Time Measurement (FTM) protocol. The FTM protocol measures the Round Trip Time (RTT) between two Wi-Fi devices which is influenced by the human-body blockage. Consequently, LocFree trains a deep classification model using the RTT data indicating the person's existence in the area. Finally, LocFree employs a smoothing stage that enables location determination with fine-grained accuracy. The evaluation of LocFree in two realistic environments demonstrates its efficacy, achieving a median localization accuracy of 1.56 meters. This surpasses the performance of state-of-the-art techniques by at least 56%. |
| 15:45 - 16:00 | Cellular Network Optimization by Deep Reinforcement Learning and AI-Enhanced Ray Tracing Zhangyu Wang, Serkan Isci, Yaron Kanza, Velin Kounev and Yusef Shaqalle Abstract: In this paper we study the use of deep reinforcement learning that is supported by a ray tracer, on top of a detailed 3D model of the geospatial environment, for optimization of antena tilts in cellular networks. We propose two novel mechanisms—geospatial importance sampling and multi-path coefficient—to efficiently pass geospatial information to the reinforcement learning model. We show that this approach can be used for fast and scalable optimization of tilt levels of cellular antennas. We present an experimental evaluation that compares the use of reinforcement learning to greedy search, simulated annealing and Bayesian optimization. Our study shows that reinforcement learning is effective and can cope with optimization problems that are at a greater scale than the settings the other algorithms can cope with. |
| 16:00 - 16:15 | Optimization of shared bicycle location in Wuhan city based on multi-source geospatial big data Shaohua Wang, Runqiao Wang, Cheng Su, Liang Zhou, Wenda Wang and Haojian Liang Abstract: With urban development and a growing focus on sustainability, shared bicycles have become a popular mode of eco-friendly transportation. However, issues like chaotic parking, excessive deployment, and suboptimal distribution need solutions. This study thoroughly analyzes how shared bicycles are used in Wuhan, focusing on time and location patterns to inform better docking point selection. It uses GPS data from Mobike shared bicycles in October 2018, point of interest (POI) data for bus stations, other facility POI data, and population distribution data for Wuhan. The research employs mathematical and statistical analysis, spatial analysis, geographical detectors, and optimization using heuristic methods and the Gurobi solver. Firstly, the study uncovers usage patterns in Wuhan, highlighting travel trends across different times and regions, and providing essential insights for selecting docking points. Secondly, it identifies the strong relationship between factors like points of interest, public transportation facilities, population density, and shared bicycle usage. The study establishes a model called the Maximum Covering Location Problem (MCLP) for selecting docking points. This model, solved using genetic algorithms and Gurobi, considers demand intensity for an optimal docking point layout. In summary, this study deepens our understanding of how shared bicycles are used in Wuhan and the significance of key factors. It creates a model for selecting docking points, offering a scientifically grounded approach and valuable insights for urban transportation planning and sustainability. |
| 16:15 - 16:20 | Closing Remarks |
Location
Radisson Blu Hotel, Congresspl. 2, 20355, Hamburg, Germany
Call For Papers (PDF version)
The workshop seeks high-quality regular (8-10 pages) and short (4 pages) papers that have not been published in other academic outlets and are not concurrently under peer review. Interested participants should submit a paper in the ACM format. Once accepted, at least one author is required to register for the workshop and the ACM SIGSPATIAL conference, as well as attend the workshop to present the accepted work which will then appear in the ACM Digital Library.
The topics include but are not limited to (in the context of industrial or related problems, such as delivery, routing, recommendation, mapping, resource allocation, and more):
Applications of AI
Applications of big data systems
Problems and benchmark datasets
Machine learning and deep learning
Computer vision and Earth observation
Generative models and simulation
Map generation
Heterogeneous data
Small data learning
Citizen science and data collection
Spatial query processing
Spatial data management and integration
Perspectives
Emerging topics and trends
Important Dates
Submission deadline
September 15, 2023 (anywhere on earth, extended)
Author notification
October 6, 2023 (anywhere on earth)
Workshop date
November 13, 2023