Frequency-Domain Diffusion Attack Framework FMDiffWA Breaches AI Watermark Defenses
Researchers propose the FMDiffWA framework, leveraging frequency-domain modulation diffusion techniques to launch highly…
📂 Research
Latest articles in Research
CharTide: Tri-Perspective Tuning Cracks the Chart-to-Code Challenge
A research team introduces the CharTide framework, which systematically addresses the core bottleneck of coupled visual …
UniME: A New Paradigm for Brain Tumor Segmentation Under Missing Modalities
Researchers propose UniME, a two-stage heterogeneous architecture that synergizes a uni-encoder with multi-encoders to e…
ArchSym: Detecting 3D Architectural Symmetry from a Single Photo
A latest arXiv paper introduces ArchSym, the first method to detect 3D symmetry structures in architectural scenes from …
EvFlow-GS: Fusing Event Cameras and Optical Flow for Motion-Deblurred 3D Reconstruction
Researchers propose the EvFlow-GS framework, which combines event camera streams with optical flow estimation to effecti…
Unsupervised Framework Breaks Through Retinal OCT Anomaly Detection Bottleneck
A research team has proposed an anatomy-aware unsupervised anomaly detection framework that can automatically detect and…
New Breakthrough in Transformer-Based Interactive Human Motion Generation
A new study proposes using Transformer models to learn and generate reactive human motions from paired interaction data,…
GenMatter: Perceiving Physical Objects with Generative Matter Models
A research team has proposed the GenMatter framework, which draws on principles of human visual perception to unify obje…
Optical Priors Unlock Novel Category Discovery in SAR Imagery
A latest arXiv paper proposes a spectrum-guided knowledge transfer method that leverages optical priors to tackle the cr…
PAGaS: Pixel-Aligned 1DoF Gaussian Splatting for Depth Refinement
Researchers propose PAGaS, a method that adapts Gaussian Splatting from novel view synthesis to depth refinement tasks. …
New Method for Large-Scale Multi-Camera Motion Capture Calibration Unveiled
A latest arXiv paper proposes a robust camera-to-mocap extrinsic calibration and verification method, addressing the pai…
FLARE-BO: A Bayesian Optimization-Driven Approach to Low-Light Vision Enhancement for Robots
A latest arXiv paper proposes FLARE-BO, a method that combines luminance enhancement with adaptive Retinex technology. B…