Computer Vision

LucidDrag: Semantic-aware dragging transforms image editing with an intention reasoner and collaborative guidance, achieving superior accuracy, image fidelity, and semantic diversity.

LiteVAE: A new autoencoder design for latent diffusion models boosts efficiency sixfold without sacrificing image quality, achieving faster training and lower memory needs via the 2D discrete wavelet …

LION: Linear Group RNNs conquer 3D object detection in sparse point clouds by enabling efficient long-range feature interaction, significantly outperforming transformer-based methods.

LinNet: A linear-time point cloud network achieving 10x speedup over PointNeXt, with state-of-the-art accuracy on various benchmarks.

Linearly decompose & recompose Vision Transformers to create diverse-scale models efficiently, reducing computational costs & improving flexibility for various applications.

Lightweight Frequency Masker significantly improves cross-domain few-shot semantic segmentation by cleverly filtering frequency components of images, thereby reducing inter-channel correlation and enh…

LE3D: Real-time HDR view synthesis from noisy RAW images is achieved using 3DGS, significantly reducing training time and improving rendering speed.

LG-CAV: Train any Concept Activation Vector with Language Guidance, leverages vision-language models to train CAVs without labeled data, achieving superior accuracy and enabling state-of-the-art model…

Lexicon3D: a first comprehensive study probing diverse visual foundation models for superior 3D scene understanding, revealing that unsupervised image models outperform others across various tasks.

ProMaC leverages MLLM hallucinations in an iterative framework to generate precise prompts for accurate object segmentation, minimizing manual prompt dependency.

Learning-to-Cache (L2C) dramatically accelerates diffusion transformers by intelligently caching layer computations, achieving significant speedups with minimal performance loss.

Meta-learning a hypernetwork on CutMix-augmented data enables data-efficient and precise correction of vision transformer errors by identifying optimal parameters for fine-tuning.

TURTLE: a novel video restoration framework that learns a truncated causal history model for efficient and high-performing video restoration, achieving state-of-the-art results on various benchmark ta…

STRAINER: A new framework enabling faster, higher-quality INR fitting by leveraging transferable features across similar signals, significantly boosting INR performance.

Decoupled Token Embedding for Merging (DTEM) significantly improves Vision Transformer efficiency by using a decoupled embedding module for relaxed token merging, achieving consistent performance gain…

AI learns to edit visual programs more accurately using a self-supervised method that combines one-shot program generation with iterative local edits, significantly boosting performance, especially wi…

Researchers developed Light Decoupling, a novel approach to model 3D facial textures under complex illumination, achieving more realistic and accurate results by decoupling unnatural lighting into mul…

Learned Mixture Density Particle Smoother (MDPS) surpasses state-of-the-art for accurate, differentiable city-scale vehicle localization.

HypStructure boosts representation learning by embedding label hierarchies into hyperbolic space, improving accuracy and interpretability.

Learned optimal lattice vector quantization (OLVQ) drastically boosts neural image compression efficiency by adapting quantizer structures to latent feature distributions, achieving significant rate-d…