Explainable AI for Time Series via Virtual Inspection Layers
For time series data, where the input itself is often not interpretable, dedicated XAI research is scarce. In this work, we put forward a virtual inspection layer for transforming the time series to an interpretable representation and allows to...
UL-DL Duality for Cell-Free Massive MIMO With Per-AP Power and Information Constraints
This article advances the theoretical foundations of user-centric cell-free massive MIMO networks. In particular, by means of a novel UL-DL duality principle for fading channels, it settles the optimality of the recently developed “team MMSE”...
Towards an AI-enabled Connected Industry: AGV Communication and Sensor Measurement Datasets
We present iV2V and iV2i+, two machine-learning datasets for industrial wireless communication. The datasets cover sidelink and cellular communication involving autonomous robots together with localization and sensing data, which can be used to...
From Hope to Safety: Unlearning Biases of Deep Models via Gradient Penalization in Latent Space
We present a novel method ensuring the right reasons on the concept level by reducing the model's sensitivity towards biases through the gradient. When modeling biases via Concept Activation Vectors, we highlight the importance of choosing robust...
Sparse Aperiodic Optical Phased Arrays on Polymer Integration Platform
Solid-state optical beam-steering utilizing polymer waveguides as edge emitters to form optical phased arrays (OPAs) with aperiodic spacing for operation at 1550 nm is demonstrated for the first time. Power consumption of 1.28 mW/? per channel is...
Animating NeRFs from Texture Space: A Framework for Pose-Dependent Rendering of Human Performances
We introduce a novel NeRF-based framework for pose-dependent rendering of human performances where the radiance field is warped around an SMPL body mesh, thereby creating a new surface-aligned representation. Our representation can be animated...
Efficient and Accurate Hyperspectral Image Demosaicing with Neural Network Architectures
This study investigates the effectiveness of neural network architectures in hyperspectral image demosaicing. We introduce a range of network models and modifications, and compare them with classical interpolation methods and existing reference...
Multi-View Inversion for 3D-aware Generative Adversarial Networks
Our method builds on existing state-of-the-art 3D GAN inversion techniques to allow for consistent and simultaneous inversion of multiple views of the same subject. We employ a multi-latent extension to handle inconsistencies present in dynamic...
Generative Texture Super-Resolution via Differential Rendering
We propose a generative deep learning network for texture map super-resolution using a differentiable renderer and calibrated reference images. Combining a super-resolution generative adversarial network (GAN) with differentiable rendering, we...
Towards Better Morphed Face Images Without Ghosting Artifacts
We propose a method for automatic prevention of ghosting artifacts based on a pixel-wise alignment during morph generation. We evaluate our proposed method on state-of-the-art detectors and show that our morphs are harder to detect, particularly,...