In the digital age, the emergence of deepfakes and synthetic media presents a significant threat to societal and political integrity. Deepfakes based on multi-modal manipulation, such as audio-visual, are more realistic and pose a greater threat. Current multi-modal deepfake detectors are often based on the attention-based fusion of heterogeneous data streams from multiple modalities. However, the heterogeneous nature of the data (such as audio and visual signals) creates a distributional modality gap and poses a significant challenge in effective fusion and hence multi-modal deepfake detection. In this paper, we propose a novel multi-modal attention framework based on recurrent neural networks (RNNs) that leverages contextual information for audio-visual deepfake detection. The proposed approach applies attention to multi-modal multi-sequence representations and learns the contributing features among them for deepfake detection and localization. Thorough experimental validations on audio-visual deepfake datasets, namely FakeAVCeleb, AV-Deepfake1M, TVIL, and LAV-DF datasets, demonstrate the efficacy of our approach. Cross-comparison with the published studies demonstrates superior performance of our approach with an improved accuracy and precision by 3.47% and 2.05% in deepfake detection and localization, respectively. Thus, obtaining state-of-the-art performance. To facilitate reproducibility, the code and the datasets information is available at https://github.com/vcbsl/audiovisual-deepfake/.

Object recognition technologies hold the potential to support blind and low-vision people in navigating the world around them. However, the gap between benchmark performances and practical usability remains a significant challenge. This paper presents a study aimed at understanding blind users' interaction with object recognition systems for identifying and avoiding errors. Leveraging a pre-existing object recognition system, URCam, fine-tuned for our experiment, we conducted a user study involving 12 blind and low-vision participants. Through in-depth interviews and hands-on error identification tasks, we gained insights into users' experiences, challenges, and strategies for identifying errors in camera-based assistive technologies and object recognition systems. During interviews, many participants preferred independent error review, while expressing apprehension toward misrecognitions. In the error identification task, participants varied viewpoints, backgrounds, and object sizes in their images to avoid and overcome errors. Even after repeating the task, participants identified only half of the errors, and the proportion of errors identified did not significantly differ from their first attempts. Based on these insights, we offer implications for designing accessible interfaces tailored to the needs of blind and low-vision users in identifying object recognition errors.

Studying low-likelihood high-impact extreme weather events in a warming world is a significant and challenging task for current ensemble forecasting systems. While these systems presently use up to 100 members, larger ensembles could enrich the sampling of internal variability. They may capture the long tails associated with climate hazards better than traditional ensemble sizes. Due to computational constraints, it is infeasible to generate huge ensembles (comprised of 1,000-10,000 members) with traditional, physics-based numerical models. In this two-part paper, we replace traditional numerical simulations with machine learning (ML) to generate hindcasts of huge ensembles. In Part I, we construct an ensemble weather forecasting system based on Spherical Fourier Neural Operators (SFNO), and we discuss important design decisions for constructing such an ensemble. The ensemble represents model uncertainty through perturbed-parameter techniques, and it represents initial condition uncertainty through bred vectors, which sample the fastest growing modes of the forecast. Using the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (IFS) as a baseline, we develop an evaluation pipeline composed of mean, spectral, and extreme diagnostics. Using large-scale, distributed SFNOs with 1.1 billion learned parameters, we achieve calibrated probabilistic forecasts. As the trajectories of the individual members diverge, the ML ensemble mean spectra degrade with lead time, consistent with physical expectations. However, the individual ensemble members' spectra stay constant with lead time. Therefore, these members simulate realistic weather states, and the ML ensemble thus passes a crucial spectral test in the literature. The IFS and ML ensembles have similar Extreme Forecast Indices, and we show that the ML extreme weather forecasts are reliable and discriminating.

It is crucial to assess the predictive performance of a model in order to establish its practicality and relevance in real-world scenarios, particularly for high-dimensional data analysis. Among data splitting or resampling methods, cross-validation (CV) is extensively used for several tasks such as estimating the prediction error, tuning the regularization parameter, and selecting the most suitable predictive model among competing alternatives. The K-fold cross-validation is a popular CV method but its limitation is that the risk estimates are highly dependent on the partitioning of the data (for training and testing). Here, the issues regarding the reproducibility of the K-fold CV estimator is demonstrated in hypothesis testing wherein different partitions lead to notably disparate conclusions. This study presents an alternative novel predictive performance test and valid confidence intervals based on exhaustive nested cross-validation for determining the difference in prediction error between two model-fitting algorithms. A naive implementation of the exhaustive nested cross-validation is computationally costly. Here, we address concerns regarding computational complexity by devising a computationally tractable closed-form expression for the proposed cross-validation estimator using ridge regularization. Our study also investigates strategies aimed at enhancing statistical power within high-dimensional scenarios while controlling the Type I error rate. To illustrate the practical utility of our method, we apply it to an RNA sequencing study and demonstrate its effectiveness in the context of biological data analysis.

Automating anomaly detection is an open problem in many scientific fields, particularly in time-domain astronomy, where modern telescopes generate millions of alerts per night. Currently, most anomaly detection algorithms for astronomical time-series rely either on hand-crafted features or on features generated through unsupervised representation learning, coupled with standard anomaly detection algorithms. In this work, we introduce a novel approach that leverages the latent space of a neural network classifier for anomaly detection. We then propose a new method called Multi-Class Isolation Forests (MCIF), which trains separate isolation forests for each class to derive an anomaly score for an object based on its latent space representation. This approach significantly outperforms a standard isolation forest when distinct clusters exist in the latent space. Using a simulated dataset emulating the Zwicky Transient Facility (54 anomalies and 12,040 common), our anomaly detection pipeline discovered $46\pm3$ anomalies ($\sim 85\%$ recall) after following up the top 2,000 ($\sim 15\%$) ranked objects. Furthermore, our classifier-based approach outperforms or approaches the performance of other state-of-the-art anomaly detection pipelines. Our novel method demonstrates that existing and new classifiers can be effectively repurposed for real-time anomaly detection. The code used in this work, including a Python package, is publicly available, https://github.com/Rithwik-G/AstroMCAD.

Motorcycle accidents pose significant risks, particularly when riders and passengers do not wear helmets. This study evaluates the efficacy of an advanced vision-language foundation model, OWLv2, in detecting and classifying various helmet-wearing statuses of motorcycle occupants using video data. We extend the dataset provided by the CVPR AI City Challenge and employ a cascaded model approach for detection and classification tasks, integrating OWLv2 and CNN models. The results highlight the potential of zero-shot learning to address challenges arising from incomplete and biased training datasets, demonstrating the usage of such models in detecting motorcycles, helmet usage, and occupant positions under varied conditions. We have achieved an average precision of 0.5324 for helmet detection and provided precision-recall curves detailing the detection and classification performance. Despite limitations such as low-resolution data and poor visibility, our research shows promising advancements in automated vehicle safety and traffic safety enforcement systems.

Artificial Intelligence (AI) is revolutionizing various fields, including public health surveillance. In Africa, where health systems frequently encounter challenges such as limited resources, inadequate infrastructure, failed health information systems and a shortage of skilled health professionals, AI offers a transformative opportunity. This paper investigates the applications of AI in public health surveillance across the continent, presenting successful case studies and examining the benefits, opportunities, and challenges of implementing AI technologies in African healthcare settings. Our paper highlights AI's potential to enhance disease monitoring and health outcomes, and support effective public health interventions. The findings presented in the paper demonstrate that AI can significantly improve the accuracy and timeliness of disease detection and prediction, optimize resource allocation, and facilitate targeted public health strategies. Additionally, our paper identified key barriers to the widespread adoption of AI in African public health systems and proposed actionable recommendations to overcome these challenges.

Touch surfaces are widely utilized for smartphones, tablet PCs, and laptops (touchpad), and single and double taps are the most basic and common operations on them. The detection of single or double taps causes the single-tap latency problem, which creates a bottleneck in terms of the sensitivity of touch inputs. To reduce the single-tap latency, we propose a novel machine-learning-based tap prediction method called PredicTaps. Our method predicts whether a detected tap is a single tap or the first contact of a double tap without having to wait for the hundreds of milliseconds conventionally required. We present three evaluations and one user evaluation that demonstrate its broad applicability and usability for various tap situations on two form factors (touchpad and smartphone). The results showed PredicTaps reduces the single-tap latency from 150-500 ms to 12 ms on laptops and to 17.6 ms on smartphones without reducing usability.

State-of-the-art task-oriented dialogue systems typically rely on task-specific ontologies for fulfilling user queries. The majority of task-oriented dialogue data, such as customer service recordings, comes without ontology and annotation. Such ontologies are normally built manually, limiting the application of specialised systems. Dialogue ontology construction is an approach for automating that process and typically consists of two steps: term extraction and relation extraction. In this work, we focus on relation extraction in a transfer learning set-up. To improve the generalisation, we propose an extension to the decoding mechanism of large language models. We adapt Chain-of-Thought (CoT) decoding, recently developed for reasoning problems, to generative relation extraction. Here, we generate multiple branches in the decoding space and select the relations based on a confidence threshold. By constraining the decoding to ontology terms and relations, we aim to decrease the risk of hallucination. We conduct extensive experimentation on two widely used datasets and find improvements in performance on target ontology for source fine-tuned and one-shot prompted large language models.

The presence of gas pores in metal feedstock powder for additive manufacturing greatly affects the final AM product. Since current porosity analysis often involves lengthy X-ray computed tomography (XCT) scans with a full rotation around the sample, motivation exists to explore methods that allow for high throughput -- possibly enabling in-line porosity analysis during manufacturing. Through labelling pore pixels on single 2D radiographs of powders, this work seeks to simulate such future efficient setups. High segmentation accuracy is achieved by combining a model of X-ray attenuation through particles with a variant of the widely applied UNet architecture; notably, F1-score increases by $11.4\%$ compared to the baseline UNet. The proposed pore segmentation is enabled by: 1) pretraining on synthetic data, 2) making tight particle cutouts, and 3) subtracting an ideal particle without pores generated from a distance map inspired by Lambert-Beers law. This paper explores four image processing methods, where the fastest (yet still unoptimized) segments a particle in mean $0.014s$ time with F1-score $0.78$, and the most accurate in $0.291s$ with F1-score $0.87$. Due to their scalable nature, these strategies can be involved in making high throughput porosity analysis of metal feedstock powder for additive manufacturing.