Long, Z., McCreadiem, R., & Imran, M. (2023). CrisisViT: A Robust Vision Transformer for Crisis Image Classification. In Jaziar Radianti, Ioannis Dokas, Nicolas Lalone, & Deepak Khazanchi (Eds.), Proceedings of the 20th International ISCRAM Conference (pp. 309–319). Omaha, USA: University of Nebraska at Omaha.
Abstract: In times of emergency, crisis response agencies need to quickly and accurately assess the situation on the ground in order to deploy relevant services and resources. However, authorities often have to make decisions based on limited information, as data on affected regions can be scarce until local response services can provide first-hand reports. Fortunately, the widespread availability of smartphones with high-quality cameras has made citizen journalism through social media a valuable source of information for crisis responders. However, analyzing the large volume of images posted by citizens requires more time and effort than is typically available. To address this issue, this paper proposes the use of state-of-the-art deep neural models for automatic image classification/tagging, specifically by adapting transformer-based architectures for crisis image classification (CrisisViT). We leverage the new Incidents1M crisis image dataset to develop a range of new transformer-based image classification models. Through experimentation over the standard Crisis image benchmark dataset, we demonstrate that the CrisisViT models significantly outperform previous approaches in emergency type, image relevance, humanitarian category, and damage severity classification. Additionally, we show that the new Incidents1M dataset can further augment the CrisisViT models resulting in an additional 1.25% absolute accuracy gain.
|
|
Gkika, I., Pattas, D., Konstantoudakis, K., & Zarpalas, D. (2023). Object detection and augmented reality annotations for increased situational awareness in light smoke conditions. In Jaziar Radianti, Ioannis Dokas, Nicolas Lalone, & Deepak Khazanchi (Eds.), Proceedings of the 20th International ISCRAM Conference (pp. 231–241). Omaha, USA: University of Nebraska at Omaha.
Abstract: Innovative technologies powered by Computer Vision algorithms can aid first responders, increasing their situ ational awareness. However, adverse conditions, such as smoke, can reduce the efficacy of such algorithms by degrading the input images. This paper presents a pipeline of image de-smoking, object detection, and augmented reality display that aims to enhance situational awareness in smoky conditions. A novel smoke-reducing deep learning algorithm is applied as a preprocessing step, before state-of-the-art object detection. The detected objects and persons are highlighted in the user’s augmented reality display. The proposed method is shown to increase detection accuracy and confidence. Testing in realistic environments provides an initial evaluation of the method, both in terms of image processing and of usefulness to first responders.
|
|
Zijun Long, & Richard McCreadie. (2022). Is Multi-Modal Data Key for Crisis Content Categorization on Social Media? In Rob Grace, & Hossein Baharmand (Eds.), ISCRAM 2022 Conference Proceedings – 19th International Conference on Information Systems for Crisis Response and Management (pp. 1068–1080). Tarbes, France.
Abstract: The user-base of social media platforms, like Twitter, has grown dramatically around the world over the last decade. As people post everything they experience on social media, large volumes of valuable multimedia content are being recorded online, which can be analysed to help for a range of tasks. Here we specifically focus on crisis response. The majority of prior works in this space focus on using machine learning to categorize single-modality content (e.g. text of the posts, or images shared), with few works jointly utilizing multiple modalities. Hence, in this paper, we examine to what extent integrating multiple modalities is important for crisis content categorization. In particular, we design a pipeline for multi-modal learning that fuses textual and visual inputs, leverages both, and then classifies that content based on the specified task. Through evaluation using the CrisisMMD dataset, we demonstrate that effective automatic labelling for this task is possible, with an average of 88.31% F1 performance across two significant tasks (relevance and humanitarian category classification). while also analysing cases that unimodal models and multi-modal models success and fail.
|
|
Nilani Algiriyage, Rangana Sampath, Raj Prasanna, Kristin Stock, Emma Hudson-Doyle, & David Johnston. (2021). Identifying Disaster-related Tweets: A Large-Scale Detection Model Comparison. In Anouck Adrot, Rob Grace, Kathleen Moore, & Christopher W. Zobel (Eds.), ISCRAM 2021 Conference Proceedings – 18th International Conference on Information Systems for Crisis Response and Management (pp. 731–743). Blacksburg, VA (USA): Virginia Tech.
Abstract: Social media applications such as Twitter and Facebook are fast becoming a key instrument in gaining situational awareness (understanding the bigger picture of the situation) during disasters. This has provided multiple opportunities to gather relevant information in a timely manner to improve disaster response. In recent years, identifying crisis-related social media posts is analysed as an automatic task using machine learning (ML) or deep learning (DL) techniques. However, such supervised learning algorithms require labelled training data in the early hours of a crisis. Recently, multiple manually labelled disaster-related open-source twitter datasets have been released. In this work, we create a large dataset with 186,718 tweets by combining a number of such datasets and evaluate the performance of multiple ML and DL algorithms in classifying disaster-related tweets in three settings, namely ``in-disaster'', ``out-disaster'' and ``cross-disaster''. Our results show that the Bidirectional LSTM model with Word2Vec embeddings performs well for the tweet classification task in all three settings. We also make available the preprocessing steps and trained weights for future research.
|
|
Zijun Long, & Richard Mccreadie. (2021). Automated Crisis Content Categorization for COVID-19 Tweet Streams. In Anouck Adrot, Rob Grace, Kathleen Moore, & Christopher W. Zobel (Eds.), ISCRAM 2021 Conference Proceedings – 18th International Conference on Information Systems for Crisis Response and Management (pp. 667–678). Blacksburg, VA (USA): Virginia Tech.
Abstract: Social media platforms, like Twitter, are increasingly used by billions of people internationally to share information. As such, these platforms contain vast volumes of real-time multimedia content about the world, which could be invaluable for a range of tasks such as incident tracking, damage estimation during disasters, insurance risk estimation, and more. By mining this real-time data, there are substantial economic benefits, as well as opportunities to save lives. Currently, the COVID-19 pandemic is attacking societies at an unprecedented speed and scale, forming an important use-case for social media analysis. However, the amount of information during such crisis events is vast and information normally exists in unstructured and multiple formats, making manual analysis very time consuming. Hence, in this paper, we examine how to extract valuable information from tweets related to COVID-19 automatically. For 12 geographical locations, we experiment with supervised approaches for labelling tweets into 7 crisis categories, as well as investigated automatic priority estimation, using both classical and deep learned approaches. Through evaluation using the TREC-IS 2020 COVID-19 datasets, we demonstrated that effective automatic labelling for this task is possible with an average of 61% F1 performance across crisis categories, while also analysing key factors that affect model performance and model generalizability across locations.
|
|