A5008517143- Advanced Vision and Imaging
- Islamic Studies and History
- Robotics and Sensor-Based Localization
- Computer Graphics and Visualization Techniques
- Video Surveillance and Tracking Methods
- Linguistic, Cultural, and Literary Studies
- Autonomous Vehicle Technology and Safety
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- 3D Shape Modeling and Analysis
- Education and Islamic Studies
- Human Pose and Action Recognition
- Generative Adversarial Networks and Image Synthesis
- Middle East Politics and Society
- Middle East and Rwanda Conflicts
- Historical and Linguistic Studies
- Remote Sensing and LiDAR Applications
- Fire Detection and Safety Systems
- Postcolonial and Cultural Literary Studies
- Global Maritime and Colonial Histories
- Advanced Image Processing Techniques
- Traffic and Road Safety
- Socioeconomic Development in MENA
- Domain Adaptation and Few-Shot Learning
- Historical Architecture and Urbanism
Menoufia University
2024
Cairo University
2023
Google (United States)
2020-2022
University of Freiburg
2016-2020
University of California, Davis
2011-2019
Columbia University
2008
We present a learning-based method for synthesizing novel views of complex scenes using only unstructured collections in-the-wild photographs. build on Neural Radiance Fields (NeRF), which uses the weights multi-layer perceptron to model density and color scene as function 3D coordinates. While NeRF works well images static subjects captured under controlled settings, it is incapable modeling many ubiquitous, real-world phenomena in uncontrolled images, such variable illumination or...
Neural Radiance Fields (NeRF) have emerged as a powerful representation for the task of novel view synthesis due to their simplicity and state-of-the-art performance. Though NeRF can produce photorealistic renderings unseen viewpoints when many input views are available, its performance drops significantly this number is reduced. We observe that majority artifacts in sparse scenarios caused by errors estimated scene geometry, divergent behavior at start training. address regularizing...
Localization is an indispensable component of a robot's autonomy stack that enables it to determine where in the environment, essentially making precursor for any action execution or planning. Although convolutional neural networks have shown promising results visual localization, they are still grossly outperformed by state-of-the-art local feature-based techniques. In this work, we propose VLocNet, new network architecture 6-DoF global pose regression and odometry estimation from...
Semantic understanding and localization are fundamental enablers of robot autonomy that have for the most part been tackled as disjoint problems. While deep learning has enabled recent breakthroughs across a wide spectrum scene tasks, its applicability to state estimation tasks limited due direct formulation renders it incapable encoding scene-specific constrains. In this work, we propose VLocNet++ architecture employs multitask approach exploit inter-task relationship between semantics,...
Data is the driving force of machine learning, with amount and quality training data often being more important for performance a system than architecture details. But collecting, processing annotating real at scale difficult, expensive, frequently raises additional privacy, fairness legal concerns. Synthetic powerful tool potential to address these shortcomings: 1) it cheap 2) supports rich ground-truth annotations 3) offers full control over 4) can circumvent or mitigate problems regarding...
A classical problem in computer vision is to infer a 3D scene representation from few images that can be used render novel views at interactive rates. Previous work focuses on reconstructing pre-defined representations, e.g. textured meshes, or implicit radiance fields, and often requires input with precise camera poses long processing times for each scene. In this work, we propose the Scene Representation Transformer (SRT), method which processes posed unposed RGB of new area, infers...
We present NeSF, a method for producing 3D semantic fields from posed RGB images alone. In place of classical representations, our builds on recent work in implicit neural scene representations wherein structure is captured by point-wise functions. leverage this methodology to recover density upon which we then train segmentation model supervised 2D maps. Despite being trained signals alone, able generate 3D-consistent maps novel camera poses and can be queried at arbitrary points. Notably,...
Text is one of the richest sources information in an urban environment. Although textual heavily relied on by humans for a majority daily tasks, its usage has not been completely exploited field robotics. In this work, we propose localization approach utilizing features environments. Starting at unknown location, equipped with RGB-camera and compass, our uses off-the-shelf text extraction methods to identify labels vicinity. We then apply probabilistic specific sensor models integrate...
For mobile robots navigating on sidewalks, the ability to safely cross street intersections is essential. Most existing approaches rely recognition of traffic light signal make an informed crossing decision. Although these have been crucial enablers for urban navigation, capabilities employing such are still limited only streets that contain signalized intersections. In this article, we address challenge and propose a multimodal convolutional neural network framework predict safety...
For mobile robots navigating on sidewalks, it is essential to be able safely cross street intersections. Most existing approaches rely the recognition of traffic light signal make an informed crossing decision. Although these have been crucial enablers for urban navigation, capabilities employing such are still limited only streets containing signalized In this paper, we address challenge and propose a multimodal convolutional neural network framework predict safety intersection crossing....
Compared to LiDAR-based localization methods, which provide high accuracy but rely on expensive sensors, visual approaches only require a camera and thus are more cost-effective while their reliability typically is inferior methods. In this work, we propose vision-based approach that learns from methods by using output as training data, combining cheap, passive sensor with an on-par localization. The consists of two deep networks trained odometry topological localization, respectively,...
We consider the problem of developing robots that navigate like pedestrians on sidewalks through city centers for performing various tasks including delivery and surveillance. One particular challenge such is crossing streets without pedestrian traffic lights. To solve this task robot has to decide based its sensory input if road clear. In work, we propose a novel multi-modal learning approach autonomous street crossing. Our solely relies laser radar data learns classifier Random Forests...
Aim: Multiple sclerosis (MS) is an autoimmune disease with a controversial etiology.Both genetic and environmental factors are thought to be involved in the risk of developing disease.The purpose this study was assess association Vitamin D receptor (VDR) BsmI variant MS investigate interaction vitamin levels.Method: 100 subjects were recruited for study.Fifty patients diagnosed 50 healthy individuals.BsmI genotyped by polymerase chain reaction (PCR) followed restriction fragment length...
Data is the driving force of machine learning, with amount and quality training data often being more important for performance a system than architecture details. But collecting, processing annotating real at scale difficult, expensive, frequently raises additional privacy, fairness legal concerns. Synthetic powerful tool potential to address these shortcomings: 1) it cheap 2) supports rich ground-truth annotations 3) offers full control over 4) can circumvent or mitigate problems regarding...
The early twentieth century witnessed the birth of an Egyptian novel that was offspring European realism and a constituent project “modernizing” country region. Yet by middle century, majority writers were veering away from in search more subjective as well indigenous or even hybrid narrative forms. seventies, however, mandated another generic shift. As impoverishing effects global capital development government's neoliberal policies became preponderant, reoriented their writings to address...
Background/Objective: D-dimer, a soluble fibrin degradation product, is used to be marker of vascular thrombosis. However, it has been reported elevated in different pathological conditions other than Moreover, its pattern post-liver transplantation (LT) children not known. So, we aimed report within the first-month post-LT and level early complications. Methods: It retrospective observational cohort study which 52 who underwent living-related liver (LRLT) were included. All available...