Sebastian Bosse

Reduan Achtibat, Maximilian Dreyer, Ilona Eisenbraun et al.

The field of eXplainable Artificial Intelligence (XAI) aims to bring transparency to today's powerful but opaque deep learning models. While local XAI methods explain individual predictions in form of attribution maps, thereby identifying where important features occur (but not providing information about what they represent), global explanation techniques visualize what concepts a model has generally learned to encode. Both types of methods thus only provide partial insights and leave the burden of interpreting the model's reasoning to the user. In this work we introduce the Concept Relevance Propagation (CRP) approach, which combines the local and global perspectives and thus allows answering both the "where" and "what" questions for individual predictions. We demonstrate the capability of our method in various settings, showcasing that CRP leads to more human interpretable explanations and provides deep insights into the model's representation and reasoning through concept atlases, concept composition analyses, and quantitative investigations of concept subspaces and their role in fine-grained decision making.

Karam Dawoud, Wojciech Samek, Peter Eisert et al.

In the ever-evolving field of Artificial Intelligence, a critical challenge has been to decipher the decision-making processes within the so-called "black boxes" in deep learning. Over recent years, a plethora of methods have emerged, dedicated to explaining decisions across diverse tasks. Particularly in tasks like image classification, these methods typically identify and emphasize the pivotal pixels that most influence a classifier's prediction. Interestingly, this approach mirrors human behavior: when asked to explain our rationale for classifying an image, we often point to the most salient features or aspects. Capitalizing on this parallel, our research embarked on a user-centric study. We sought to objectively measure the interpretability of three leading explanation methods: (1) Prototypical Part Network, (2) Occlusion, and (3) Layer-wise Relevance Propagation. Intriguingly, our results highlight that while the regions spotlighted by these methods can vary widely, they all offer humans a nearly equivalent depth of understanding. This enables users to discern and categorize images efficiently, reinforcing the value of these methods in enhancing AI transparency.

Michael Gerstenberger, Steffen Maaß, Peter Eisert et al.

We introduce a Gaussian Prototype Layer for gradient-based prototype learning and demonstrate two novel network architectures for explainable segmentation one of which relies on region proposals. Both models are evaluated on agricultural datasets. While Gaussian Mixture Models (GMMs) have been used to model latent distributions of neural networks before, they are typically fitted using the EM algorithm. Instead, the proposed prototype layer relies on gradient-based optimization and hence allows for end-to-end training. This facilitates development and allows to use the full potential of a trainable deep feature extractor. We show that it can be used as a novel building block for explainable neural networks. We employ our Gaussian Prototype Layer in (1) a model where prototypes are detected in the latent grid and (2) a model inspired by Fast-RCNN with SLIC superpixels as region proposals. The earlier achieves a similar performance as compared to the state-of-the art while the latter has the benefit of a more precise prototype localization that comes at the cost of slightly lower accuracies. By introducing a gradient-based GMM layer we combine the benefits of end-to-end training with the simplicity and theoretical foundation of GMMs which will allow to adapt existing semi-supervised learning strategies for prototypical part models in future.

Ari Wahl, Dorian Gawlinski, David Przewozny et al.

Pre-trained general-purpose Vision-Language Models (VLM) hold the potential to enhance intuitive human-machine interactions due to their rich world knowledge and 2D object detection capabilities. However, VLMs for 3D coordinates detection tasks are rare. In this work, we investigate interactive abilities of VLMs by returning 3D object positions given a monocular RGB image from a wrist-mounted camera, natural language input, and robot states. We collected and curated a heterogeneous dataset of more than 100,000 images and finetuned a VLM using QLoRA with a custom regression head. By implementing conditional routing, our model maintains its ability to process general visual queries while adding specialized 3D position estimation capabilities. Our results demonstrate robust predictive performance with a median MAE of 13 mm on the test set and a five-fold improvement over a simpler baseline without finetuning. In about 25% of the cases, predictions are within a range considered acceptable for the robot to interact with objects.

Michael Gerstenberger, Sebastian Lapuschkin, Peter Eisert et al.

Despite significant advances in machine learning, decision-making of artificial agents is still not perfect and often requires post-hoc human interventions. If the prediction of a model relies on unreasonable factors it is desirable to remove their effect. Deep interactive prototype adjustment enables the user to give hints and correct the model's reasoning. In this paper, we demonstrate that prototypical-part models are well suited for this task as their prediction is based on prototypical image patches that can be interpreted semantically by the user. It shows that even correct classifications can rely on unreasonable prototypes that result from confounding variables in a dataset. Hence, we propose simple yet effective interaction schemes for inference adjustment: The user is consulted interactively to identify faulty prototypes. Non-object prototypes can be removed by prototype masking or a custom mode of deselection training. Interactive prototype rejection allows machine learning naïve users to adjust the logic of reasoning without compromising the accuracy.

Birgit Nierula, Karam Tomotaki-Dawoud, Daniel Johannes Meyer et al.

We present the early-stage design and implementation of a multimodal, real-time communication analysis system intended as a foundational interaction layer for adaptive VR training. The system integrates five parallel processing streams: (1) verbal and prosodic speech analysis, (2) skeletal gesture recognition from multi-view RGB cameras, (3) multimodal affective analysis combining lower-face video with upper-face facial EMG, (4) EEG-based mental state decoding, and (5) physiological arousal estimation from skin conductance, heart activity, and proxemic behavior. All signals are synchronized via Lab Streaming Layer to enable temporally aligned, continuous assessments of users' conscious and unconscious communication cues. Building on concepts from social semiotics and symbolic interactionism, we introduce an interpretation layer that links low-level signal representations to interactional constructs such as escalation and de-escalation. This layer is informed by domain knowledge from police instructors and lay participants, grounding system responses in realistic conflict scenarios. We demonstrate the feasibility and limitations of automated cue extraction in an XR-based de-escalation training project for law enforcement, reporting preliminary results for gesture recognition, emotion recognition under HMD occlusion, verbal assessment, mental state decoding, and physiological arousal. Our findings highlight the value of multi-view sensing and multimodal fusion for overcoming occlusion and viewpoint challenges, while underscoring that fusion and feedback must be treated as design problems rather than purely technical ones. The work contributes design resources and empirical insights for shaping human-AI-powered XR training in complex interpersonal settings.

Katerina Katsarou, George Zountsas, Karam Tomotaki-Dawoud et al.

Reliable monitoring of surgical instrument exchanges is essential for maintaining procedural efficiency and patient safety in the operating room. Automatic detection of instrument handovers in intraoperative video remains challenging due to frequent occlusions, background clutter, and the temporally evolving nature of interaction events. We propose a spatiotemporal vision framework for event-level detection and direction classification of surgical instrument handovers in surgical videos. The model combines a Vision Transformer (ViT) backbone for spatial feature extraction with a unidirectional Long Short-Term Memory (LSTM) network for temporal aggregation. A unified multi-task formulation jointly predicts handover occurrence and interaction direction, enabling consistent modeling of transfer dynamics while avoiding error propagation typical of cascaded pipelines. Predicted confidence scores form a temporal signal over the video, from which discrete handover events are identified via peak detection. Experiments on a dataset of kidney transplant procedures demonstrate strong performance, achieving an F1-score of 0.84 for handover detection and a mean F1-score of 0.72 for direction classification, outperforming both a single-task variant and a VideoMamba-based baseline for direction prediction while maintaining comparable detection performance. To improve interpretability, we employ Layer-CAM attribution to visualize spatial regions driving model decisions, highlighting hand-instrument interaction cues.

Laura-Sophia von Hirschhausen, Jannes S. Magnusson, Mykyta Kovalenko et al.

Deep learning has transformed computer vision for precision agriculture, yet apple orchard monitoring remains limited by dataset constraints. The lack of diverse, realistic datasets and the difficulty of annotating dense, heterogeneous scenes. Existing datasets overlook different growth stages and stereo imagery, both essential for realistic 3D modeling of orchards and tasks like fruit localization, yield estimation, and structural analysis. To address these gaps, we present AppleGrowthVision, a large-scale dataset comprising two subsets. The first includes 9,317 high resolution stereo images collected from a farm in Brandenburg (Germany), covering six agriculturally validated growth stages over a full growth cycle. The second subset consists of 1,125 densely annotated images from the same farm in Brandenburg and one in Pillnitz (Germany), containing a total of 31,084 apple labels. AppleGrowthVision provides stereo-image data with agriculturally validated growth stages, enabling precise phenological analysis and 3D reconstructions. Extending MinneApple with our data improves YOLOv8 performance by 7.69 % in terms of F1-score, while adding it to MinneApple and MAD boosts Faster R-CNN F1-score by 31.06 %. Additionally, six BBCH stages were predicted with over 95 % accuracy using VGG16, ResNet152, DenseNet201, and MobileNetv2. AppleGrowthVision bridges the gap between agricultural science and computer vision, by enabling the development of robust models for fruit detection, growth modeling, and 3D analysis in precision agriculture. Future work includes improving annotation, enhancing 3D reconstruction, and extending multimodal analysis across all growth stages.

Arian Beckmann, Tilman Stephani, Felix Klotzsche et al.

Since the advent of Deepfakes in digital media, the development of robust and reliable detection mechanism is urgently called for. In this study, we explore a novel approach to Deepfake detection by utilizing electroencephalography (EEG) measured from the neural processing of a human participant who viewed and categorized Deepfake stimuli from the FaceForensics++ datset. These measurements serve as input features to a binary support vector classifier, trained to discriminate between real and manipulated facial images. We examine whether EEG data can inform Deepfake detection and also if it can provide a generalized representation capable of identifying Deepfakes beyond the training domain. Our preliminary results indicate that human neural processing signals can be successfully integrated into Deepfake detection frameworks and hint at the potential for a generalized neural representation of artifacts in computer generated faces. Moreover, our study provides next steps towards the understanding of how digital realism is embedded in the human cognitive system, possibly enabling the development of more realistic digital avatars in the future.

Sören Becker, Thomas Wiegand, Sebastian Bosse

The performance of visual quality prediction models is commonly assumed to be closely tied to their ability to capture perceptually relevant image aspects. Models are thus either based on sophisticated feature extractors carefully designed from extensive domain knowledge or optimized through feature learning. In contrast to this, we find feature extractors constructed from random noise to be sufficient to learn a linear regression model whose quality predictions reach high correlations with human visual quality ratings, on par with a model with learned features. We analyze this curious result and show that besides the quality of feature extractors also their quantity plays a crucial role - with top performances only being achieved in highly overparameterized models.

Serhan Gül, Sebastian Bosse, Dimitri Podborski et al.

Volumetric video allows viewers to experience highly-realistic 3D content with six degrees of freedom in mixed reality (MR) environments. Rendering complex volumetric videos can require a prohibitively high amount of computational power for mobile devices. A promising technique to reduce the computational burden on mobile devices is to perform the rendering at a cloud server. However, cloud-based rendering systems suffer from an increased interaction (motion-to-photon) latency that may cause registration errors in MR environments. One way of reducing the effective latency is to predict the viewer's head pose and render the corresponding view from the volumetric video in advance. In this paper, we design a Kalman filter for head motion prediction in our cloud-based volumetric video streaming system. We analyze the performance of our approach using recorded head motion traces and compare its performance to an autoregression model for different prediction intervals (look-ahead times). Our results show that the Kalman filter can predict head orientations 0.5 degrees more accurately than the autoregression model for a look-ahead time of 60 ms.

Andrew Perkis, Christian Timmerer, Sabina Baraković et al.

With the coming of age of virtual/augmented reality and interactive media, numerous definitions, frameworks, and models of immersion have emerged across different fields ranging from computer graphics to literary works. Immersion is oftentimes used interchangeably with presence as both concepts are closely related. However, there are noticeable interdisciplinary differences regarding definitions, scope, and constituents that are required to be addressed so that a coherent understanding of the concepts can be achieved. Such consensus is vital for paving the directionality of the future of immersive media experiences (IMEx) and all related matters. The aim of this white paper is to provide a survey of definitions of immersion and presence which leads to a definition of immersive media experience (IMEx). The Quality of Experience (QoE) for immersive media is described by establishing a relationship between the concepts of QoE and IMEx followed by application areas of immersive media experience. Influencing factors on immersive media experience are elaborated as well as the assessment of immersive media experience. Finally, standardization activities related to IMEx are highlighted and the white paper is concluded with an outlook related to future developments.

Sebastian Bosse, Dominique Maniry, Klaus-Robert Müller et al.

We present a deep neural network-based approach to image quality assessment (IQA). The network is trained end-to-end and comprises ten convolutional layers and five pooling layers for feature extraction, and two fully connected layers for regression, which makes it significantly deeper than related IQA models. Unique features of the proposed architecture are that: 1) with slight adaptations it can be used in a no-reference (NR) as well as in a full-reference (FR) IQA setting and 2) it allows for joint learning of local quality and local weights, i.e., relative importance of local quality to the global quality estimate, in an unified framework. Our approach is purely data-driven and does not rely on hand-crafted features or other types of prior domain knowledge about the human visual system or image statistics. We evaluate the proposed approach on the LIVE, CISQ, and TID2013 databases as well as the LIVE In the wild image quality challenge database and show superior performance to state-of-the-art NR and FR IQA methods. Finally, cross-database evaluation shows a high ability to generalize between different databases, indicating a high robustness of the learned features.

Rafael Reisenhofer, Sebastian Bosse, Gitta Kutyniok et al.

In most practical situations, the compression or transmission of images and videos creates distortions that will eventually be perceived by a human observer. Vice versa, image and video restoration techniques, such as inpainting or denoising, aim to enhance the quality of experience of human viewers. Correctly assessing the similarity between an image and an undistorted reference image as subjectively experienced by a human viewer can thus lead to significant improvements in any transmission, compression, or restoration system. This paper introduces the Haar wavelet-based perceptual similarity index (HaarPSI), a novel and computationally inexpensive similarity measure for full reference image quality assessment. The HaarPSI utilizes the coefficients obtained from a Haar wavelet decomposition to assess local similarities between two images, as well as the relative importance of image areas. The consistency of the HaarPSI with the human quality of experience was validated on four large benchmark databases containing thousands of differently distorted images. On these databases, the HaarPSI achieves higher correlations with human opinion scores than state-of-the-art full reference similarity measures like the structural similarity index (SSIM), the feature similarity index (FSIM), and the visual saliency-based index (VSI). Along with the simple computational structure and the short execution time, these experimental results suggest a high applicability of the HaarPSI in real world tasks.