A5114889359- Generative Adversarial Networks and Image Synthesis
- Advanced Vision and Imaging
- Computer Graphics and Visualization Techniques
- Metallurgy and Material Forming
- Advanced Neural Network Applications
- Image and Video Stabilization
- Handwritten Text Recognition Techniques
- Human Motion and Animation
- Nanofabrication and Lithography Techniques
- 3D Shape Modeling and Analysis
- Video Surveillance and Tracking Methods
- Model Reduction and Neural Networks
- Industrial Vision Systems and Defect Detection
- Advanced Image and Video Retrieval Techniques
- Gaussian Processes and Bayesian Inference
- Video Analysis and Summarization
University of Science and Technology of China
2023-2024
One tough problem of image inpainting is to restore complex structures in the corrupted regions. It motivates interactive which leverages additional hints, e.g., sketches, assist process. A sketch simple and intuitive for end users provide, but meanwhile has free forms with much randomness. Such randomness may confuse models, incur severe artifacts completed images. To better facilitate guidance, we propose a two-stage system, termed SketchRefiner. The first stage our approach serves as data...
Despite the promising performance of current video segmentation models on existing benchmarks, these still struggle with complex scenes. In this paper, we introduce 6th Large-scale Video Object Segmentation (LSVOS) challenge in conjunction ECCV 2024 workshop. This year's includes two tasks: (VOS) and Referring (RVOS). year, replace classic YouTube-VOS YouTube-RVOS benchmark latest datasets MOSE, LVOS, MeViS to assess VOS under more challenging environments. attracted 129 registered teams...
Recent advancements of generative AI have significantly promoted content creation and editing, where prevailing studies further extend this exciting progress to video editing. In doing so, these mainly transfer the inherent motion patterns from source videos edited ones, results with inferior consistency user prompts are often observed, due lack particular alignments between delivered motions contents. To address limitation, we present a shape-consistent editing method, namely StableV2V, in...
Recent advances in NeRF inpainting have leveraged pretrained diffusion models to enhance performance. However, these methods often yield suboptimal results due their ineffective utilization of 2D priors. The limitations manifest two critical aspects: the inadequate capture geometric information by and guidance provided existing Score Distillation Sampling (SDS) methods. To address problems, we introduce GB-NeRF, a novel framework that enhances through improved Our approach incorporates key...
This paper introduces UnZipLoRA, a method for decomposing an image into its constituent subject and style, represented as two distinct LoRAs (Low-Rank Adaptations). Unlike existing personalization techniques that focus on either or style in isolation, require separate training sets each, UnZipLoRA disentangles these elements from single by both the simultaneously. ensures resulting are compatible, i.e., they can be seamlessly combined using direct addition. enables independent manipulation...
Promotional videos are rapidly becoming a popular medium for persuading people to change their behaviours in many settings (e.g., online shopping, social enterprise initiatives). Today, such often produced by professionals, which is time-, labour- and cost-intensive undertaking. In order produce contents support large applications e-commerce), the field of artificial intelligence (AI)-empowered persuasive video generation (AIPVG) has gained traction recent years. This interdisciplinary...
One tough problem of image inpainting is to restore complex structures in the corrupted regions. It motivates interactive which leverages additional hints, e.g., sketches, assist process. Sketch simple and intuitive end users, but meanwhile has free forms with much randomness. Such randomness may confuse models, incur severe artifacts completed images. To address this problem, we propose a two-stage method termed SketchRefiner. In first stage, using cross-correlation loss function robustly...
In recent years, generative adversarial networks (GANs) have been an actively studied topic and shown to successfully produce high-quality realistic images in various domains. The controllable synthesis ability of GAN generators suggests that they maintain informative, disentangled, explainable image representations, but leveraging transferring their representations downstream tasks is largely unexplored. this paper, we propose distill knowledge from by squeezing spanning representations. We...