Labeling data instances is an important task in machine learning and visual analytics. Both fields provide a broad set of labeling strategies, whereby (and particular active learning) follows rather model-centered approach analytics employs user-centered approaches (visual-interactive labeling). have individual strengths weaknesses. In this work, we conduct experiment with three parts to assess compare the performance these different strategies. our study, (1) identify strategies for...

Recently enacted legislation grants individuals certain rights to decide in what fashion their personal data may be used and particular a "right forgotten". This poses challenge machine learning: how proceed when an individual retracts permission use which has been part of the training process model? From this question emerges field unlearning, could broadly described as investigation "delete from models". Our work complements direction research for specific setting class-wide deletion...

Abstract The quantification of ground reaction forces (GRF) is a standard tool for clinicians to quantify and analyze human locomotion. Such recordings produce vast amount complex data variables which are difficult comprehend. This makes interpretation challenging. Machine learning approaches seem be promising tools support in identifying categorizing specific gait patterns. However, the quality such strongly depends on available annotated train underlying models. Therefore, we present G ait...

Machine Learning (ML) is increasingly used to support decision-making in the healthcare sector. While ML approaches provide promising results with regard their classification performance, most share a central limitation, black-box character. This article investigates usefulness of Explainable Artificial Intelligence (XAI) methods increase transparency automated clinical gait based on time series. For this purpose, predictions state-of-the-art are explained XAI method called Layer-wise...

Recent comparative data reveal that formant frequencies are cues to body size in animals, due a close relationship between frequency spacing, vocal tract length and overall size. Accordingly, intriguing morphological adaptations elongate the order lower formants occur several species, with exaggeration hypothesis being proposed justify most of these observations. While elephant trunk is strongly implicated account for low rumbles, it unknown whether elephants emit vocalizations exclusively...

Biologists often have to investigate large amounts of video in behavioral studies animals. These videos are usually not sufficiently indexed which makes the finding objects interest a time-consuming task. We propose fully automated method for detection and tracking elephants wildlife has been collected by biologists field. The dynamically learns color model from few training images. Based on model, we localize sequences with different backgrounds lighting conditions. exploit temporal clues...

Abstract The labeling of data sets is a time‐consuming task, which is, however, an important prerequisite for machine learning and visual analytics. Visual‐interactive (VIAL) provides users active role in the process labeling, with goal to combine potentials humans machines make more efficient. Recent experiments showed that apply different strategies when selecting instances visual‐interactive interfaces. In this paper, we contribute systematic quantitative analysis such user strategies. We...

The human-elephant conflict is one of the most serious conservation problems in Asia and Africa today. involuntary confrontation humans elephants claims lives many animals every year. A promising approach to alleviate this development an acoustic early warning system. Such a system requires robust automated detection elephant vocalizations under unconstrained field conditions. Today, no exists that fulfills these requirements. In paper, we present method for diverse noise sources field. We...

The protection of private information is a crucial issue in data-driven research and business contexts. Typically, techniques like anonymisation or (selective) deletion are introduced order to allow data sharing, e. g. the case collaborative endeavours. For use with techniques, $k$-anonymity criterion one most popular, numerous scientific publications on different algorithms metrics. Anonymisation often require changing thus necessarily affect results machine learning models trained...

Human gait is a complex and unique biological process that can offer valuable insights into an individual's health well-being. In this work, we leverage machine learning-based approach to model individual signatures identify factors contributing inter-individual variability in patterns. We provide comprehensive analysis of individuality by (1) demonstrating the uniqueness large-scale dataset (2) highlighting characteristics are most distinctive each individual. utilized data from three...

In 2014, more than 10 million people in the US were affected by an ambulatory disability. Thus, gait rehabilitation is a crucial part of health care systems. The quantification human locomotion enables clinicians to describe and analyze patient's performance detail allows them base clinical decisions on objective data. These assessments generate vast amount complex data which need be interpreted short time period. We conducted design study cooperation with analysis experts develop novel...

Petroglyphs (rock engravings) have been pecked and engraved by humans into natural rock surfaces thousands of years ago are among the oldest artifacts that document early human life culture. Some these engravings survived until present serve today as a unique ancient life. Since petroglyphs surface rocks, they threatened environmental factors such weather erosion. To preserve valuable history, 3D digitization has become suitable approach due to development powerful reconstruction techniques...

Until recently few research has been performed in the area of animal sound retrieval. The authors identify state-of-the-art techniques general purpose recognition by a broad survey literature. Based on findings, this paper gives thorough investigation audio features and classifiers their applicability domain sounds. We introduce set novel descriptors compare quality to other popular features. results are encouraging motivate further

Animal vocal signals are increasingly used to monitor wildlife populations and obtain estimates of species occurrence abundance. In the future, acoustic monitoring should function not only detect animals, but also extract detailed information about by discriminating sexes, age groups, social or kin potentially individuals. Here we show that it is possible estimate groups African elephants (Loxodonta africana) based on parameters extracted from rumbles recorded under field conditions in a...

This article proposes a comprehensive investigation of the automatic classification functional gait disorders based solely on ground reaction force (GRF) measurements. The aim study is twofold: (1) to investigate suitability stateof-the-art GRF parameterization techniques (representations) for discrimination disorders; and (2) provide first performance baseline automated large-scale dataset. utilized database comprises measurements from 279 patients with (GDs) data 161 healthy controls (N)....

This paper presents our contribution to the ChaLearn Challenge 2015 on Cultural Event Classification. The challenge in this task is automatically classify images from 50 different cultural events. Our solution based combination of visual features extracted convolutional neural networks with temporal information using a hierarchical classifier scheme. We extract last three fully connected layers both CaffeNet (pre-trained ImageNet) and fine tuned version for challenge. propose late fusion...

The detection of a specific social event requires for high semantic understanding in the interpretation particular characteristics such as its type and location. In many cases, photos capturing different events at same (or highly similar) locations can hardly be distinguished by each other. Available metadata provide assistance where there is no expert knowledge hand. However, often lack completeness reliability. this paper, we explore feasibility fully automated approach events. comparison...

The decline of habitat for elephants due to expanding human activity is a serious conservation problem. This has continuously escalated the human-elephant conflict in Africa and Asia. Elephants make extensive use powerful infrasonic calls (rumbles) that travel distances up several kilometers. makes well-suited acoustic monitoring because it enables detecting even if they are out sight. In sight, their distinct visual appearance them good candidate monitoring. We provide an integrated...

Social media is accompanied by an increasing pro-portion of content that provides fake information or misleading content, known as disorder. In this paper, we study the problem multimodal news detection on a large-scale dataset. We propose network architecture enables different levels and types fusion. addition to textual visual posting, further leverage secondary information, i.e. user comments metadata. fuse at multiple account for specific intrinsic structure modalities. Our results show...

This work investigates the effectiveness of various machine learning (ML) methods in classifying human gait patterns associated with cerebral palsy (CP) and examines clinical relevance learned features using explainability approaches. We trained different ML models, including convolutional neural networks, self-normalizing random forests, decision trees, generated explanations for models. For deep Grad-CAM were aggregated on levels to obtain at decision, class model level. investigate which...

The automated acoustic detection of elephants is an important factor in alleviating the human-elephant conflict Asia and Africa. In this paper, we present a method for elephant presence evaluate it on large dataset wildlife recordings. We introduce novel technique signal enhancement to improve robustness detector noisy situations. Experiments show that proposed outperforms existing methods strongly improves noise sources from environment. first step towards system presence.