Computer Vision

WISE: Evaluates world knowledge in text-to-image generation.

Seedream 2.0: A native Chinese-English bilingual image generation model that understands cultural nuances and excels in text rendering.

RayFlow: Accelerating diffusion with instance-aware adaptive flow, boosting speed & quality!

PLADIS: Sparsity boosts attention for diffusion models, enhancing text-to-image generation at inference time!

PE3R: Achieves fast and accurate 3D scene reconstruction from 2D images by enhanced perception and efficiency.

eMIGM: A unified, efficient masked image generation model achieving state-of-the-art performance with fewer resources.

EasyControl: Efficient & flexible control for Diffusion Transformers, enabling sophisticated image generation.

DreamRelation: Personalize videos by customizing relationships between subjects, generalizing to new domains.

Seg-Zero: Cognitive Reinforcement for Reasoning-Chain Guided Segmentation!

TDM: a new diffusion distillation paradigm unifying trajectory distillation and distribution matching, surpassing teachers in a data-free manner with state-of-the-art performance and low training cost…

DropletVideo: A dataset and approach to explore integral spatio-temporal consistent video generation.

VideoPainter: Edit any video, any length, with user-guided instructions!

TrajectoryCrafter: Precisely control camera movement in monocular videos with a novel diffusion model for coherent 4D content generation.

MagicInfinite: Infinite talking videos from words and voice!

AnyAnomaly: LVLM for customizable zero-shot video anomaly detection, adapting to diverse environments without retraining.

ProReflow: Improves diffusion model efficiency via progressive training and direction-focused velocity alignment.

RectifiedHR: Enables training-free high-resolution image generation via energy rectification, boosting both efficiency and effectiveness.

Q-Eval-100K: A new, large dataset for evaluating visual quality and text alignment in AI-generated content.

Kiss3DGen generates 3D assets by repurposing 2D diffusion models, enabling efficient 3D editing and enhancement.

Likelihood-based generative models get a GAN-like boost via a new Direct Discriminative Optimization, ditching the joint training complexity.