Physics Maths Engineering
Ilias Gialampoukidis,
Ilias Gialampoukidis
Information Technologies Institute, Centre for Research and Technology Hellas,
Thomas Papadimos,
Thomas Papadimos
Information Technologies Institute, Centre for Research and Technology Hellas
Stelios Andreadis,
Stelios Andreadis
Information Technologies Institute, Centre for Research and Technology Hellas
Stefanos Vrochidis
Stefanos Vrochidis
Information Technologies Institute, Centre for Research and Technology Hellas
Peer Reviewed
This paper discusses the importance of detecting breaking events in real time to help emergency response workers, and how social media can be used to process large amounts of data quickly. Most event detection techniques have focused on either images or text, but combining the two can improve performance. The authors present lessons learned from the Flood-related multimedia task in MediaEval2020, provide a dataset for reproducibility, and propose a new multimodal fusion method that uses Graph Neural Networks to combine image, text, and time information. Their method outperforms state-of-the-art approaches and can handle low-sample labelled data.
Real-time event detection helps emergency responders quickly identify and react to breaking events, such as natural disasters or accidents. By analyzing social media data, responders can gain timely insights and allocate resources more effectively.
Social media platforms provide a vast amount of real-time data, including images, text, and timestamps. Analyzing this data can help detect events faster and more accurately than traditional methods, especially when combined using advanced techniques like multimodal fusion.
Multimodal fusion combines different types of data, such as images, text, and time information, to improve event detection. This approach leverages the strengths of each data type, leading to better performance than using just one type of data.
Graph Neural Networks (GNNs) are used to combine and analyze multimodal data effectively. In this study, GNNs helped integrate image, text, and time information, outperforming traditional methods and handling low-sample labeled data efficiently.
The study highlighted the importance of combining image and text data for event detection. It also demonstrated that advanced techniques like GNNs can significantly improve performance, even with limited labeled data.
The proposed method uses GNNs for multimodal fusion, which outperforms state-of-the-art approaches. It also addresses the challenge of low-sample labeled data, making it more practical for real-world applications.
The study provides a dataset from the MediaEval2020 Flood-related task, which includes social media posts with images, text, and timestamps. This dataset is made available for reproducibility and further research.
Combining images and text provides a more comprehensive view of events. For example, an image might show flood damage, while the accompanying text provides context like location and severity. Together, they offer richer information for accurate detection.
By analyzing social media data in real time, this method can quickly identify disaster-affected areas, assess damage, and provide actionable insights to emergency responders, improving response times and saving lives.
GNNs excel at handling complex relationships between different types of data, such as images, text, and time. They are also effective with limited labeled data, making them ideal for real-world event detection tasks.
Researchers can use the provided dataset and the proposed GNN-based method to reproduce and build on the study’s results. Developers can apply these techniques to create tools for real-time event detection, especially for emergency response and disaster management.
Show by month | Manuscript | Video Summary |
---|---|---|
2025 April | 2 | 2 |
2025 March | 60 | 60 |
2025 February | 48 | 48 |
2025 January | 43 | 43 |
2024 December | 45 | 45 |
2024 November | 40 | 40 |
2024 October | 58 | 58 |
2024 September | 44 | 44 |
2024 August | 31 | 31 |
2024 July | 33 | 33 |
2024 June | 20 | 20 |
2024 May | 38 | 38 |
2024 April | 30 | 30 |
2024 March | 8 | 8 |
Total | 500 | 500 |
Show by month | Manuscript | Video Summary |
---|---|---|
2025 April | 2 | 2 |
2025 March | 60 | 60 |
2025 February | 48 | 48 |
2025 January | 43 | 43 |
2024 December | 45 | 45 |
2024 November | 40 | 40 |
2024 October | 58 | 58 |
2024 September | 44 | 44 |
2024 August | 31 | 31 |
2024 July | 33 | 33 |
2024 June | 20 | 20 |
2024 May | 38 | 38 |
2024 April | 30 | 30 |
2024 March | 8 | 8 |
Total | 500 | 500 |