A5054694693- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Reservoir Engineering and Simulation Methods
- Computer Graphics and Visualization Techniques
- 3D Shape Modeling and Analysis
- Hydraulic Fracturing and Reservoir Analysis
- Industrial Vision Systems and Defect Detection
- Image Enhancement Techniques
- Image Processing Techniques and Applications
- Manufacturing Process and Optimization
- Enhanced Oil Recovery Techniques
- Generative Adversarial Networks and Image Synthesis
- Oil and Gas Production Techniques
- 3D Surveying and Cultural Heritage
- Advanced Image Fusion Techniques
- Medical Image Segmentation Techniques
- Advanced Optical Sensing Technologies
- Food Drying and Modeling
- CO2 Sequestration and Geologic Interactions
- Human Pose and Action Recognition
- Aluminum Alloy Microstructure Properties
- Additive Manufacturing and 3D Printing Technologies
- Remote Sensing and LiDAR Applications
- Geological and Geophysical Studies
University of Petroleum and Energy Studies
2022-2025
University Of Information Technology
2025
Wyższa Szkoła Technologii Informatycznych w Warszawie
2025
META Health
2023-2024
Petronas (Malaysia)
2014-2024
ABES Engineering College
2023-2024
Institute of Management Technology
2023-2024
Ansal University
2024
Vikram Sarabhai Space Centre
2023
Manipal Academy of Higher Education
2023
The aim of this paper is to propose a mechanism efficiently and explicitly model image hierarchies in the global, regional, local range for restoration. To achieve that, we start by analyzing two important properties natural images including cross-scale similarity anisotropic features. Inspired anchored stripe self-attention which achieves good balance between space time complexity modelling capacity beyond regional range. Then new network architecture dubbed GRL Global, Regional, Local via...
Video restoration (e.g., video super-resolution) aims to restore high-quality frames from low-quality frames. Different single image restoration, generally requires utilize temporal information multiple adjacent but usually misaligned Existing deep methods tackle with this by exploiting a sliding window strategy or recurrent architecture, which either is restricted frame-by-frame lacks long-range modelling ability. In paper, we propose Restoration Transformer (VRT) parallel frame prediction...
Video restoration aims to restore high-quality frames from low-quality frames. Different single image restoration, video generally requires utilize temporal information multiple adjacent but usually misaligned Existing deep methods tackle with this by exploiting a sliding window strategy or recurrent architecture, which are restricted frame-by-frame restoration. In paper, we propose Restoration Transformer (VRT) parallel frame prediction ability. More specifically, VRT is composed of scales,...
Video restoration aims at restoring multiple high-quality frames from low-quality frames. Existing video methods generally fall into two extreme cases, i.e., they either restore all in parallel or the frame by a recurrent way, which would result different merits and drawbacks. Typically, former has advantage of temporal information fusion. However, it suffers large model size intensive memory consumption; latter relatively small as shares parameters across frames; however, lacks long-range...
The aim of this paper is to propose a large scale dataset for image restoration (LSDIR). Recent work in has been focused on the design deep neural networks. datasets used train these networks 'only' contain some thousands images, which still incomparable with other vision tasks such as visual recognition and object detection. small training set limits performance To solve problem, we collect high-resolution (HR) images from Flickr restoration. ensure pixel-level quality collected dataset,...
Diffusion models have emerged as a powerful tool for point cloud generation. A key component that drives the impressive performance generating high-quality samples from noise is iteratively denoise thousands of steps. While beneficial, complexity learning steps has limited its applications to many 3D real-world. To address this limitation, we propose Point Straight Flow (PSF), model exhibits using one step. Our idea based on reformulation standard diffusion model, which optimizes curvy...
This investigation focuses on the wear resistance and surface morphology of multi-walled carbon nanotube (MWCNT)-filled bio-based epoxy composites. study examines impact different MWCNT concentrations (0 Wt.%, 0.25 0.50 0.75 Wt.%) properties these Techniques such as scanning electron microscopy (SEM) atomic force (AFM) were utilized for comprehensive characterization. The results demonstrated a direct correlation between content composites, which corroborated by robust statistical analysis....
Metal matrix composites (MMCs) have achieved significant attention in engineering applications because of their exceptional properties, like increased strength-to-weight ratiosand resistance to wear. However, manufacturing processes pose challenges for industries, such as oxidation, porosity, and chemical reactions. To address these challenges, this study investigates the processing sintering (500 °C) Ti-6Al-4V-SiCp mechanical particularly hardness, wear frictional force using a statistical...
In this paper, we study a novel problem in egocentric action recognition, which term as "Multimodal Generalization" (MMG). MMG aims to how systems can generalize when data from certain modalities is limited or even completely missing. We thoroughly investigate the context of standard supervised recognition and more challenging few-shot setting for learning new categories. consists two scenarios, designed support security, efficiency considerations real-world applications: (1) missing...
To apply optical flow in practice, it is often necessary to resize the input smaller dimensions order reduce computational costs. However, downsizing inputs makes estimation more challenging because objects and motion ranges become smaller. Even though recent approaches have demonstrated high-quality estimation, they tend fail accurately model small precise boundaries when resolution lowered, restricting their applicability high-resolution inputs. In this paper, we introduce AnyFlow, a...
Neural radiance fields (NeRFs) enable novel-view synthesis with unprecedented visual quality. However, to render photorealistic images, NeRFs require hundreds of deep multilayer perceptron (MLP) evaluations - for each pixel. This is prohibitively expensive and makes realtime rendering infeasible, even on powerful modern GPUs. In this paper, we propose a novel approach distill bake into highly efficient mesh-based neural representations that are fully compatible the massively parallel...
Efficiently reconstructing accurate 3D models from monocular video is a key challenge in computer vision, critical for advancing applications virtual reality, robotics, and scene understanding. Existing approaches typically require pre-computed camera parameters frame-by-frame reconstruction pipelines, which are prone to error accumulation entail significant computational overhead. To address these limitations, we introduce VideoLifter, novel framework that leverages geometric priors...
A two-dimensional finite-element formulation and solution of a set transient coupled heat diffusive moisture transfer equations is presented. The procedure developed uses an alpha family approximation for stepping in time the applied to simulate stepwise convective drying behavior banana slices. model tested was validated with experimental data from different sources using pump dryer (HPD) as well continuous batch both Cartesian cylindrical coordinate systems. maximum deviation content...
We propose an efficient neural network for RAW image denoising. Although network-based denoising has been extensively studied restoration, little attention given to compute limited and power sensitive devices, such as smartphones wearables. In this paper, we present a novel architecture suite of training techniques high quality in mobile devices. Our work is distinguished by three main contributions. (1) The Feature-Align layer that modulates the activations encoder-decoder with input noisy...
Direct time-of-flight (dToF) sensors are promising for next-generation on-device 3D sensing. However, limited by manufacturing capabilities in a compact module, the dToF data has low spatial resolution (e.g. <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim 20\times 30$</tex> iPhone dToF), and it requires super-resolution step before being passed to downstream tasks. In this paper, we solve problem fusing low-resolution with corresponding...