Thomas Altstidl
YOU?
Author Swipe
View article: Stratify or Die: Rethinking Data Splits in Image Segmentation
Stratify or Die: Rethinking Data Splits in Image Segmentation Open
Random splitting of datasets in image segmentation often leads to unrepresentative test sets, resulting in biased evaluations and poor model generalization. While stratified sampling has proven effective for addressing label distribution i…
View article: A Cross-Platform Smartphone Auscultation SDK and Optimized Filters for Severe Aortic Stenosis Detection
A Cross-Platform Smartphone Auscultation SDK and Optimized Filters for Severe Aortic Stenosis Detection Open
Initial studies suggest that valve replacement may also benefit asymptomatic patients with severe aortic stenosis, who don't typically seek medical attention and thus require screening. As echocardiography, the current gold standard, is ti…
View article: Caption-Driven Explorations: Aligning Image and Text Embeddings through Human-Inspired Foveated Vision
Caption-Driven Explorations: Aligning Image and Text Embeddings through Human-Inspired Foveated Vision Open
Understanding human attention is crucial for vision science and AI. While many models exist for free-viewing, less is known about task-driven image exploration. To address this, we introduce CapMIT1003, a dataset with captions and click-co…
View article: How Intermodal Interaction Affects the Performance of Deep Multimodal Fusion for Mixed-Type Time Series
How Intermodal Interaction Affects the Performance of Deep Multimodal Fusion for Mixed-Type Time Series Open
Mixed-type time series (MTTS) is a bimodal data type that is common in many domains, such as healthcare, finance, environmental monitoring, and social media. It consists of regularly sampled continuous time series and irregularly sampled c…
View article: Large-Scale Dataset Pruning in Adversarial Training through Data Importance Extrapolation
Large-Scale Dataset Pruning in Adversarial Training through Data Importance Extrapolation Open
Their vulnerability to small, imperceptible attacks limits the adoption of deep learning models to real-world systems. Adversarial training has proven to be one of the most promising strategies against these attacks, at the expense of a su…
View article: The Impact of Random Models on Standardized Clustering Similarity
The Impact of Random Models on Standardized Clustering Similarity Open
Clustering similarity measures are essential for evaluating clustering results and ensuring diversity in multiple clusterings of the same dataset. Common indices like the Mutual Information (MI) and Rand Index (RI) are biased towards small…
View article: Detection of aortic stenosis using built-in microphones of commercially available smartphones
Detection of aortic stenosis using built-in microphones of commercially available smartphones Open
Background According to recent data, asymptomatic patients with severe aortic stenosis benefit from early treatment. Hence, widely applicable methods to identify patients with severe aortic stenosis in an asymptomatic state may be desirabl…
View article: Contrastive Language-Image Pretrained Models are Zero-Shot Human Scanpath Predictors
Contrastive Language-Image Pretrained Models are Zero-Shot Human Scanpath Predictors Open
Understanding the mechanisms underlying human attention is a fundamental challenge for both vision science and artificial intelligence. While numerous computational models of free-viewing have been proposed, less is known about the mechani…
View article: Raising the Bar for Certified Adversarial Robustness with Diffusion Models
Raising the Bar for Certified Adversarial Robustness with Diffusion Models Open
Certified defenses against adversarial attacks offer formal guarantees on the robustness of a model, making them more reliable than empirical methods such as adversarial training, whose effectiveness is often later reduced by unseen attack…
View article: Scaled and Translated Image Recognition (STIR) Source Data
Scaled and Translated Image Recognition (STIR) Source Data Open
While convolutions are known to be invariant to (discrete) translations, scaling continues to be a challenge and most image recognition networks are not invariant to them. To explore these effects, we have created the Scaled and Translated…
View article: Scaled and Translated Image Recognition (STIR) Source Data
Scaled and Translated Image Recognition (STIR) Source Data Open
While convolutions are known to be invariant to (discrete) translations, scaling continues to be a challenge and most image recognition networks are not invariant to them. To explore these effects, we have created the Scaled and Translated…
View article: Just a Matter of Scale? Reevaluating Scale Equivariance in Convolutional Neural Networks
Just a Matter of Scale? Reevaluating Scale Equivariance in Convolutional Neural Networks Open
The widespread success of convolutional neural networks may largely be attributed to their intrinsic property of translation equivariance. However, convolutions are not equivariant to variations in scale and fail to generalize to objects o…
View article: Scaled and Translated Image Recognition (STIR)
Scaled and Translated Image Recognition (STIR) Open
Paper: [2211.10288] Just a Matter of Scale? Reevaluating Scale Equivariance in Convolutional Neural Networks (arxiv.org) Code: taltstidl/scale-equivariant-cnn: Official code for "Just a Matter of Scale? Reevaluating Scale Equivariance in C…
View article: Scaled and Translated Image Recognition (STIR)
Scaled and Translated Image Recognition (STIR) Open
Paper: [2211.10288] Just a Matter of Scale? Reevaluating Scale Equivariance in Convolutional Neural Networks (arxiv.org) Code: taltstidl/scale-equivariant-cnn: Official code for "Just a Matter of Scale? Reevaluating Scale Equivariance in C…