arXiv subjects

Mohamed Amer

Publications and source records attributed to Mohamed Amer.

5 recordsLinked to original sources

arXiv2025

Unmanned Aerial Vehicles Control in a Digital Twin: Exploring the Effect of Different Points of View on User Experience in Virtual Reality

Controlling Unmanned Aerial Vehicles (UAVs) is a cognitively demanding task, with accidents often arising from insufficient situational awareness, inadequate training, and poor user experiences. Providing more intuitive and immersive visual feedback, particularly through Digital Twin technologies, offers new opportunities to enhance pilot awareness and overall experience quality. In this study, we investigate how different virtual points of view (POVs) influence user experience and performance during UAV piloting in Virtual Reality (VR), utilizing a digital twin that faithfully replicates the real-world flight environment. We developed a VR application that enables participants to control a physical DJI Mini 4 Pro drone while immersed in a digital twin with four distinct camera perspectives: Baseline View (static external), First-Person View, Chase View, and Third-Person View. Nineteen participants completed a series of ring-based obstacle courses from each perspective. In addition to objective flight data, we collected standardized subjective assessments of user experience, presence, workload, cybersickness, and situational awareness. Quantitative analyses revealed that the First-Person View was associated with significantly higher mental demand and effort, greater trajectory deviation, but smoother control inputs compared to the Third-Person and Chase perspectives. Complementing these findings, preference data indicated that the Third-Person View was most consistently favored, whereas the First-Person View elicited polarized reactions.

cs.HC↗

arXiv2025

Finding My Way: Influence of Different Audio Augmented Reality Navigation Cues on User Experience and Subjective Usefulness

As augmented reality (AR) becomes increasingly prevalent in mobile and context-aware applications, the role of auditory cues in guiding users through physical environments is becoming critical. This study investigates the effectiveness and user experience of various categories of audio cues, including fully non-verbal sounds and speech-derived Spearcons, during outdoor navigation tasks using the Meta Quest 3 headset. Twenty participants navigated five outdoor routes using audio-only cue types: Artificial Sounds, Nature Sounds, Spearcons, Musical Instruments, and Auditory Icons. Subjective evaluations were collected to assess the perceived effectiveness and user experience of each sound type. Results revealed significant differences in perceived novelty and stimulation across sound types. Artificial Sounds and Musical Instruments were rated higher than Spearcons in novelty, while Artificial Sounds were also rated higher than Spearcons in stimulation. Overall preference was evenly split between Nature Sounds and Artificial Sounds. These findings suggest that incorporating aspects of novelty and user engagement in auditory feedback design may enhance the effectiveness of AR navigation systems.

cs.HC↗

arXiv2018

BitNet: Bit-Regularized Deep Neural Networks

We present a novel optimization strategy for training neural networks which we call "BitNet". The parameters of neural networks are usually unconstrained and have a dynamic range dispersed over all real values. Our key idea is to limit the expressive power of the network by dynamically controlling the range and set of values that the parameters can take. We formulate this idea using a novel end-to-end approach that circumvents the discrete parameter space by optimizing a relaxed continuous and differentiable upper bound of the typical classification loss function. The approach can be interpreted as a regularization inspired by the Minimum Description Length (MDL) principle. For each layer of the network, our approach optimizes real-valued translation and scaling factors and arbitrary precision integer-valued parameters (weights). We empirically compare BitNet to an equivalent unregularized model on the MNIST and CIFAR-10 datasets. We show that BitNet converges faster to a superior quality solution. Additionally, the resulting model has significant savings in memory due to the use of integer-valued parameters.

cs.LG↗

arXiv2017

GPU Activity Prediction using Representation Learning

GPU activity prediction is an important and complex problem. This is due to the high level of contention among thousands of parallel threads. This problem was mostly addressed using heuristics. We propose a representation learning approach to address this problem. We model any performance metric as a temporal function of the executed instructions with the intuition that the flow of instructions can be identified as distinct activities of the code. Our experiments show high accuracy and non-trivial predictive power of representation learning on a benchmark.

cs.LG↗

arXiv2017

Low Precision Neural Networks using Subband Decomposition

Large-scale deep neural networks (DNN) have been successfully used in a number of tasks from image recognition to natural language processing. They are trained using large training sets on large models, making them computationally and memory intensive. As such, there is much interest in research development for faster training and test time. In this paper, we present a unique approach using lower precision weights for more efficient and faster training phase. We separate imagery into different frequency bands (e.g. with different information content) such that the neural net can better learn using less bits. We present this approach as a complement existing methods such as pruning network connections and encoding learning weights. We show results where this approach supports more stable learning with 2-4X reduction in precision with 17X reduction in DNN parameters.

cs.LG↗