A5008162341- Computer Graphics and Visualization Techniques
- Advanced Vision and Imaging
- Digital Media Forensic Detection
- Generative Adversarial Networks and Image Synthesis
- Speech and Audio Processing
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Multimodal Machine Learning Applications
- Epigenetics and DNA Methylation
- Industrial Vision Systems and Defect Detection
- Image Processing and 3D Reconstruction
- Advanced ceramic materials synthesis
- Advanced Data Compression Techniques
- Face recognition and analysis
- Data Management and Algorithms
- 3D Shape Modeling and Analysis
- Medical Image Segmentation Techniques
- Additive Manufacturing and 3D Printing Technologies
Fraunhofer Institute for Telecommunications, Heinrich Hertz Institute
2023-2025
Humboldt-Universität zu Berlin
2024-2025
University of Augsburg
2022-2023
Abstract 3D Gaussian Splatting (3DGS) has emerged as a cutting‐edge technique for real‐time radiance field rendering, offering state‐of‐the‐art performance in terms of both quality and speed. 3DGS models scene collection three‐dimensional Gaussians, with additional attributes optimized to conform the scene's geometric visual properties. Despite its advantages rendering speed image fidelity, is limited by significant storage memory demands. These high demands make impractical mobile devices...
Visually sorted grid layouts provide an efficient method for organizing high-dimensional vectors in two-dimensional space by aligning spatial proximity with similarity relationships. This approach facilitates the effective sorting of diverse elements ranging from data points to images, and enables simultaneous visualization a significant number elements. However, on grids is challenge due its high complexity. Even small 8-by-8 64 elements, possible arrangements exceeds 1.3 * 10^89 - more...
Current scene graph datasets suffer from strong long-tail distributions of their predicate classes. Due to a very low number some classes in the test sets, no reliable metrics can be retrieved for rarest We construct new panoptic dataset and set that are designed as benchmark predictive performance especially on rare To dataset, we propose model-assisted annotation pipeline efficiently finds hidden large images like needles haystack.Contrary prior datasets, Haystack contains explicit...
3D Gaussian Splatting has recently emerged as a highly promising technique for modeling of static scenes. In contrast to Neural Radiance Fields, it utilizes efficient rasterization allowing very fast rendering at high-quality. However, the storage size is significantly higher, which hinders practical deployment, e.g. on resource constrained devices. this paper, we introduce compact scene representation organizing parameters (3DGS) into 2D grid with local homogeneity, ensuring drastic...
NeRF-based 3D-aware Generative Adversarial Networks (GANs) like EG3D or GIRAFFE have shown very high rendering quality under large representational variety. However, with Neural Radiance Fields poses challenges for 3D applications: First, the significant computational demands of NeRF preclude its use on low-power devices, such as mobiles and VR/AR headsets. Second, implicit representations based neural networks are difficult to incorporate into explicit scenes, VR environments video games....
We present a work-in-progress survey on 3D Gaussian Splatting compression methods, focusing their statistical performance across various benchmarks. This aims to facilitate comparability by summarizing key statistics of different approaches in tabulated format. The datasets evaluated include TanksAndTemples, MipNeRF360, DeepBlending, and SyntheticNeRF. For each method, we report the Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Learned Perceptual Image Patch (LPIPS),...
Abstract In this work, we propose a method that enforces explicit control over various attributes during the image generation process in generative adversarial net. We semi-supervised learning procedure allows us to use quantized approximation of object orientation for continuous rotations. As result, among many other attributes, our proposed scenes are rendered according specifications.
Current scene graph datasets suffer from strong long-tail distributions of their predicate classes. Due to a very low number some classes in the test sets, no reliable metrics can be retrieved for rarest We construct new panoptic dataset and set that are designed as benchmark predictive performance especially on rare To dataset, we propose model-assisted annotation pipeline efficiently finds hidden large images like needles haystack. Contrary prior datasets, Haystack contains explicit...
Current 3D GAN inversion methods for human heads typically use only one single frontal image to reconstruct the whole head model. This leaves out meaningful information when multi-view data or dynamic videos are available. Our method builds on existing state-of-the-art techniques allow consistent and simultaneous of multiple views same subject. We employ a multi-latent extension handle inconsistencies present in face re-synthesize representations from sequence. As our uses additional about...