Computer Vision

MSPE empowers Vision Transformers to handle any image resolution by cleverly optimizing patch embedding, achieving superior performance on low-resolution images and comparable results on high-resoluti…

MotionGS enhances deformable 3D Gaussian splatting for dynamic scenes by using motion flow to guide deformation, significantly improving reconstruction accuracy and outperforming state-of-the-art meth…

MotionCraft: Physics-based zero-shot video generation creates realistic videos with complex motion dynamics by cleverly warping the noise latent space of an image diffusion model using optical flow fr…

Motion Graph unleashes efficient and accurate video prediction by transforming video frames into interconnected graph nodes, capturing complex motion patterns with minimal computational cost.

Boosting video diffusion: Motion Consistency Model (MCM) disentangles motion and appearance learning for high-fidelity, fast video generation using few sampling steps.

MoTE: A novel framework harmonizes generalization and specialization for visual-language video knowledge transfer, achieving state-of-the-art results.

MonoMAE enhances monocular 3D object detection by using depth-aware masked autoencoders to effectively handle object occlusions, achieving superior performance on both occluded and non-occluded object…

MonkeySee reconstructs natural images from macaque brain signals with high accuracy using a novel CNN decoder, advancing neural decoding and offering insights into visual perception.

MoE Jetpack efficiently transforms readily available dense checkpoints into high-performing MoE models, drastically accelerating convergence and improving accuracy.

Hydra: a novel cryo-EM reconstruction method resolves both conformational and compositional heterogeneity ab initio, enabling the analysis of complex, unpurified samples with state-of-the-art accuracy…

Boosting Robust Few-Shot Learning with Adversarial Meta-Tuning!

Researchers propose a simple yet effective clustering-based adaptive margin loss to mitigate biases inherited by black-box feature extractors in image classification tasks.

INTEGER: a novel unsupervised point cloud registration method leveraging feature-geometry coherence for reliable pseudo-label mining and density-invariant feature learning, achieving state-of-the-art …

CoPA improves cross-domain few-shot learning by adapting separate transformations for prototype and image embeddings, significantly enhancing performance and revealing better representation clusters.

MimicTalk generates realistic, expressive talking videos in minutes using a pre-trained model adapted to individual identities.

MIDGARD: Generate high-quality, simulatable 3D articulated assets with enhanced control and interpretability using a novel diffusion-based framework.

Humans as landmarks: A novel zero-shot monocular metric depth estimation method leverages generative models and human mesh recovery to transfer metric scale information, achieving superior generalizat…

MetaUAS achieves universal visual anomaly segmentation using only one normal image prompt via a pure vision model, surpassing previous zero-shot, few-shot, and full-shot methods.

Meta-Exploiting Frequency Prior enhances cross-domain few-shot learning by leveraging image frequency decomposition and consistency priors to improve model generalization and efficiency.

Meta 3D AssetGen: High-quality text-to-mesh generation with realistic PBR materials and lighting, exceeding prior methods in speed and accuracy.