Computer Vision

EasyRef uses multimodal LLMs to generate images from multiple references, overcoming limitations of prior methods by capturing consistent visual elements and offering improved zero-shot generalization…

DisPose disentangles pose guidance for controllable human image animation, generating diverse animations while preserving appearance consistency using only sparse skeleton pose input, eliminating the …

InvSR: a novel image super-resolution technique using diffusion inversion, enabling flexible sampling steps for efficient and high-fidelity results.

DiTFlow: training-free video motion transfer using Diffusion Transformers, enabling realistic motion control in synthesized videos via Attention Motion Flow.

UniReal: a universal framework for image generation and editing, unifying diverse tasks via learning real-world dynamics from video data, achieving highly realistic and versatile results.

STIV: A novel, scalable method for text and image-conditioned video generation, systematically improving model architectures, training, and data curation for superior performance.

OmniDocBench, a novel benchmark, tackles limitations in current document parsing by introducing a diverse, high-quality dataset with comprehensive annotations, enabling fair multi-level evaluation of …

ObjCtrl-2.5D: Training-free, precise image-to-video object control using 3D trajectories and camera poses.

MobileVD: The first mobile-optimized video diffusion model, achieving 523x efficiency improvement over state-of-the-art with minimal quality loss, enabling realistic video generation on smartphones.

FiVA dataset and its adaptation framework enable unprecedented fine-grained control over visual attributes in text-to-image generation, empowering users to craft highly customized images.

FireFlow makes editing images faster and better.

Introducing Evaluation Agent, a faster, more flexible human-like framework for evaluating visual generative AI.

DiffSensei: A new framework generates customized manga with dynamic multi-character control using multi-modal LLMs and diffusion models, outperforming existing methods.

MoViE: Mobile Diffusion for Video Editing achieves 12 FPS video editing on mobile phones by optimizing existing image editing models, achieving a major breakthrough in on-device video processing.

EMOv2 achieves state-of-the-art performance in various vision tasks using a novel Meta Mobile Block, pushing the 5M parameter lightweight model frontier.

MinT: Generating coherent videos with precisely timed, multiple events via temporal control, surpassing existing methods.

RAPL efficiently aligns robots with human preferences using minimal feedback by aligning visual representations before reward learning.

LiFT leverages human feedback, including reasoning, to effectively align text-to-video models with human preferences, significantly improving video quality.

ZipAR accelerates autoregressive image generation by up to 91% through parallel decoding leveraging spatial locality in images, making high-resolution image generation significantly faster.

SwiftEdit achieves lightning-fast, high-quality text-guided image editing in just 0.23 seconds via a novel one-step diffusion process.