Search

The softmax function is used as an activation function placed in the output layer of a neural network. It allows extracting the probabilities of the output classes, while introduces a non-linearity to the model. In the field of low-end FPGAs, implementations of Deep Neural Networks (DNNs) require the exploration of optimisation techniques to improve computational efficiency and hardware resource consumption. This work explores approximate computing techniques to implement the softmax function, using Taylor and Pad\'e approximations, and interpolation methods with Look-Up Tables (LUTs). The introduction of approximations aims to reduce the required execution time while reducing the precision of results produced by the softmax function. Each implementation is evaluated using Root Mean Square Error (RMSE) for accuracy assessment, and individual performance is verified by taking measurements of execution times. From our evaluation, quadratic interpolation with LUTs achieves the lowest error, but in terms of performance, Taylor and Pad\'e approximations show better execution times, which highlights the existing design trade-off between numerical accuracy and power consumption.

Automated Urban Air Mobility (UAM) can improve passenger transportation and reduce congestion, but its success depends on passenger trust. While initial research addresses passengers' information needs, questions remain about how to simulate air taxi flights and how these simulations impact users and interface requirements. We conducted a between-subjects study (N=40), examining the influence of motion fidelity in Virtual-Reality-simulated air taxi flights on user effects and interface design. Our study compared simulations with and without motion cues using a 3-Degrees-of-Freedom motion chair. Optimizing the interface design across six objectives, such as trust and mental demand, we used multi-objective Bayesian optimization to determine the most effective design trade-offs. Our results indicate that motion fidelity decreases users' trust, understanding, and acceptance, highlighting the need to consider motion fidelity in future UAM studies to approach realism. However, minimal evidence was found for differences or equality in the optimized interface designs, suggesting personalized interface designs.

Infinite scrolling on social media platforms is designed to encourage prolonged engagement, leading users to spend more time than desired, which can provoke negative emotions. Interventions to mitigate infinite scrolling have shown initial success, yet users become desensitized due to the lack of contextual relevance. Understanding how contextual factors influence intervention effectiveness remains underexplored. We conducted a 7-day user study (N=72) investigating how these contextual factors affect users' reactance and responsiveness to interventions during infinite scrolling. Our study revealed an interplay, with contextual factors such as being at home, sleepiness, and valence playing significant roles in the intervention's effectiveness. Low valence coupled with being at home slows down the responsiveness to interventions, and sleepiness lowers reactance towards interventions, increasing user acceptance of the intervention. Overall, our work contributes to a deeper understanding of user responses toward interventions and paves the way for developing more effective interventions during infinite scrolling.

The introduction of Highly Automated Vehicles (HAVs) has the potential to increase the independence of blind and visually impaired people (BVIPs). However, ensuring safety and situation awareness when exiting these vehicles in unfamiliar environments remains challenging. To address this, we conducted an interactive workshop with N=5 BVIPs to identify their information needs when exiting an HAV and evaluated three prior-developed low-fidelity prototypes. The insights from this workshop guided the development of PathFinder, a multimodal interface combining visual, auditory, and tactile modalities tailored to BVIP's unique needs. In a three-factorial within-between-subject study with N=16 BVIPs, we evaluated PathFinder against an auditory-only baseline in urban and rural scenarios. PathFinder significantly reduced mental demand and maintained high perceived safety in both scenarios, while the auditory baseline led to lower perceived safety in the urban scenario compared to the rural one. Qualitative feedback further supported PathFinder's effectiveness in providing spatial orientation during exiting.

Object perception from multi-view cameras is crucial for intelligent systems, particularly in indoor environments, e.g., warehouses, retail stores, and hospitals. Most traditional multi-target multi-camera (MTMC) detection and tracking methods rely on 2D object detection, single-view multi-object tracking (MOT), and cross-view re-identification (ReID) techniques, without properly handling important 3D information by multi-view image aggregation. In this paper, we propose a 3D object detection and tracking framework, named BEV-SUSHI, which first aggregates multi-view images with necessary camera calibration parameters to obtain 3D object detections in bird's-eye view (BEV). Then, we introduce hierarchical graph neural networks (GNNs) to track these 3D detections in BEV for MTMC tracking results. Unlike existing methods, BEV-SUSHI has impressive generalizability across different scenes and diverse camera settings, with exceptional capability for long-term association handling. As a result, our proposed BEV-SUSHI establishes the new state-of-the-art on the AICity'24 dataset with 81.22 HOTA, and 95.6 IDF1 on the WildTrack dataset.

Quantum magnetometers based on spin defects in solids enable sensitive imaging of various magnetic phenomena, such as ferro- and antiferromagnetism, superconductivity, and current-induced fields. Existing protocols primarily focus on static fields or narrow-band dynamical signals, and are optimized for high sensitivity rather than fast time resolution. Here, we report detection of fast signal transients, providing a perspective for investigating the rich dynamics of magnetic systems. We experimentally demonstrate our technique using a single nitrogen-vacancy (NV) center magnetometer at room temperature, reaching a best-effort time resolution of 1.1 ns, an instantaneous bandwidth of 0.9 GHz, and a time-of-flight precision better than 20 ps. The time resolution can be extended to the picosecond range by use of on-chip waveguides. At these speeds, NV quantum magnetometers will become competitive with time-resolved synchrotron X-ray techniques. Looking forward, adding fast temporal resolution to the spatial imaging capability further promotes single-spin probes as powerful research tools in spintronics, mesoscopic physics, and nanoscale device metrology., Comment: 11 pages including methods

In the recent paradigm of Federated Learning (FL), multiple clients train a shared model while keeping their local data private. Resource constraints of clients and communication costs pose major problems for training large models in FL. On the one hand, addressing the resource limitations of the clients, sparse training has proven to be a powerful tool in the centralized setting. On the other hand, communication costs in FL can be addressed by local training, where each client takes multiple gradient steps on its local data. Recent work has shown that local training can provably achieve the optimal accelerated communication complexity [Mishchenko et al., 2022]. Hence, one would like an accelerated sparse training algorithm. In this work we show that naive integration of sparse training and acceleration at the server fails, and how to fix it by letting the clients perform these tasks appropriately. We introduce Sparse-ProxSkip, our method developed for the nonconvex setting, inspired by RandProx [Condat and Richt\'arik, 2022], which provably combines sparse training and acceleration in the convex setting. We demonstrate the good performance of Sparse-ProxSkip in extensive experiments.

The largest direct dark matter search experiments to date employ dual-phase time projection chambers (TPCs) with liquid noble gas targets. These detect both the primary photons generated by particle interactions in the liquid target, as well as proportional secondary scintillation light created by the ionization electrons in a strong electric field in the gas phase between the liquid-gas interface and the anode. In this work, we describe the detection of charge signals in a small-scale single-phase liquid-xenon-only TPC, that features the well-established TPC geometry with light readout above and below a cylindrical target. In the single-phase TPC, the proportional scintillation light (S2) is generated in liquid xenon in close proximity to 10 {\mu}m diameter anode wires. The detector was characterized and the proportional scintillation process was studied using the 32.1 keV and 9.4 keV signals from 83mKr decays. A charge gain factor g2 of up to (1.9 $\pm$ 0.3) PE/electron was reached at an anode voltage 4.4 kV higher than the gate electrode 5 mm below it, corresponding to (29 $\pm$ 6) photons emitted per ionization electron. The duration of S2 signals is dominated by electron diffusion and approaches the xenon de-excitation timescale for very short electron drift times. The electron drift velocity and the longitudinal diffusion constant were measured at a drift field of 470 V/cm. The results agree with the literature and demonstrate that a single-phase TPC can be operated successfully., Comment: 20 pages, 14 figures

We propose a novel method for geolocalizing Unmanned Aerial Vehicles (UAVs) in environments lacking Global Navigation Satellite Systems (GNSS). Current state-of-the-art techniques employ an offline-trained encoder to generate a vector representation (embedding) of the UAV's current view, which is then compared with pre-computed embeddings of geo-referenced images to determine the UAV's position. Here, we demonstrate that the performance of these methods can be significantly enhanced by preprocessing the images to extract their edges, which exhibit robustness to seasonal and illumination variations. Furthermore, we establish that utilizing edges enhances resilience to orientation and altitude inaccuracies. Additionally, we introduce a confidence criterion for localization. Our findings are substantiated through synthetic experiments., Comment: Accepted for publication in IGARSS2024. 4 pages, 3 figures, 3 tables

We propose the SAL (Segment Anything in Lidar) method consisting of a text-promptable zero-shot model for segmenting and classifying any object in Lidar, and a pseudo-labeling engine that facilitates model training without manual supervision. While the established paradigm for Lidar Panoptic Segmentation (LPS) relies on manual supervision for a handful of object classes defined a priori, we utilize 2D vision foundation models to generate 3D supervision ``for free''. Our pseudo-labels consist of instance masks and corresponding CLIP tokens, which we lift to Lidar using calibrated multi-modal data. By training our model on these labels, we distill the 2D foundation models into our Lidar SAL model. Even without manual labels, our model achieves $91\%$ in terms of class-agnostic segmentation and $54\%$ in terms of zero-shot Lidar Panoptic Segmentation of the fully supervised state-of-the-art. Furthermore, we outperform several baselines that do not distill but only lift image features to 3D. More importantly, we demonstrate that SAL supports arbitrary class prompts, can be easily extended to new datasets, and shows significant potential to improve with increasing amounts of self-labeled data. Code and models are available at this $\href{https://github.com/nv-dvl/segment-anything-lidar}{URL}$., Comment: Accepted to ECCV 2024

Federated Learning (FL) has garnered increasing attention due to its unique characteristic of allowing heterogeneous clients to process their private data locally and interact with a central server, while being respectful of privacy. A critical bottleneck in FL is the communication cost. A pivotal strategy to mitigate this burden is \emph{Local Training}, which involves running multiple local stochastic gradient descent iterations between communication phases. Our work is inspired by the innovative \emph{Scaffnew} algorithm, which has considerably advanced the reduction of communication complexity in FL. We introduce FedComLoc (Federated Compressed and Local Training), integrating practical and effective compression into \emph{Scaffnew} to further enhance communication efficiency. Extensive experiments, using the popular TopK compressor and quantization, demonstrate its prowess in substantially reducing communication overheads in heterogeneous settings.

Until recently, the Video Instance Segmentation (VIS) community operated under the common belief that offline methods are generally superior to a frame by frame online processing. However, the recent success of online methods questions this belief, in particular, for challenging and long video sequences. We understand this work as a rebuttal of those recent observations and an appeal to the community to focus on dedicated near-online VIS approaches. To support our argument, we present a detailed analysis on different processing paradigms and the new end-to-end trainable NOVIS (Near-Online Video Instance Segmentation) method. Our transformer-based model directly predicts spatio-temporal mask volumes for clips of frames and performs instance tracking between clips via overlap embeddings. NOVIS represents the first near-online VIS approach which avoids any handcrafted tracking heuristics. We outperform all existing VIS methods by large margins and provide new state-of-the-art results on both YouTube-VIS (2019/2021) and the OVIS benchmarks.

The advent of data-driven technology solutions is accompanied by an increasing concern with data privacy. This is of particular importance for human-centered image recognition tasks, such as pedestrian detection, re-identification, and tracking. To highlight the importance of privacy issues and motivate future research, we motivate and introduce the Pedestrian Dataset De-Identification (PDI) task. PDI evaluates the degree of de-identification and downstream task training performance for a given de-identification method. As a first baseline, we propose IncogniMOT, a two-stage full-body de-identification pipeline based on image synthesis via generative adversarial networks. The first stage replaces target pedestrians with synthetic identities. To improve downstream task performance, we then apply stage two, which blends and adapts the synthetic image parts into the data. To demonstrate the effectiveness of IncogniMOT, we generate a fully de-identified version of the MOT17 pedestrian tracking dataset and analyze its application as training data for pedestrian re-identification, detection, and tracking models. Furthermore, we show how our data is able to narrow the synthetic-to-real performance gap in a privacy-conscious manner.

While supervised neural networks have become state of the art for identifying the rare strong gravitational lenses from large imaging data sets, their selection remains significantly affected by the large number and diversity of nonlens contaminants. This work evaluates and compares systematically the performance of neural networks in order to move towards a rapid selection of galaxy-scale strong lenses with minimal human input in the era of deep, wide-scale surveys. We used multiband images from PDR2 of the HSC Wide survey to build test sets mimicking an actual classification experiment, with 189 strong lenses previously found over the HSC footprint and 70,910 nonlens galaxies in COSMOS. Multiple networks were trained on different sets of realistic strong-lens simulations and nonlens galaxies, with various architectures and data pre-processing. The overall performances strongly depend on the construction of the ground-truth training data and they typically, but not systematically, improve using our baseline residual network architecture. Improvements are found when applying random shifts to the image centroids and square root stretches to the pixel values, adding z band, or using random viewpoints of the original images, but not when adding difference images to subtract emission from the central galaxy. The most significant gain is obtained with committees of networks trained on different data sets, and showing a moderate overlap between populations of false positives. Nearly-perfect invariance to image quality can be achieved by training networks either with large number of bands, or jointly with the PSF and science frames. Overall, we show the possibility to reach a TPR0 as high as 60% for the test sets under consideration, which opens promising perspectives for pure selection of strong lenses without human input using the Rubin Observatory and other forthcoming ground-based surveys., Comment: 21 pages, 10 figures, submitted to A&A, comments are welcome

This position paper discusses the negative effects of excessive smartphone usage on mental health and well-being. Despite efforts to limit smartphone usage, users become desensitized to reminders and limitations. The paper proposes the use of Push & Pull Factors to contextualize intervention strategies promoting digital well-being. Further, alternative metrics to measure the effectiveness of intervention strategies will be discussed., Comment: 8 pages