Image Generation

ART: A novel automatic red-teaming framework reveals safety vulnerabilities in popular text-to-image models by identifying unsafe outputs even from seemingly harmless prompts.

Boosting image generation: Applying guidance selectively during diffusion model sampling drastically enhances image quality and inference speed, achieving state-of-the-art results.

AnyFit: Controllable virtual try-on for any attire combination across any scenario, exceeding existing methods in accuracy and scalability.

Image generation gets a speed boost with TiTok, a novel 1D image tokenizer that uses just 32 tokens for high-quality image reconstruction and generation, achieving up to 410x faster processing than st…

EMDiffusion trains clean diffusion models from corrupted data using an expectation-maximization algorithm, achieving state-of-the-art results on diverse imaging tasks.

DiMR boosts image generation fidelity by cleverly combining multi-resolution networks with time-dependent layer normalization in diffusion models, achieving state-of-the-art results on ImageNet.

Diffusion-KTO: Aligning text-to-image models with human preferences using simple likes/dislikes, maximizing expected human utility.

AirSketch generates aesthetically pleasing sketches directly from noisy hand-motion tracking data using a self-supervised controllable diffusion model, eliminating the need for expensive AR/VR equipme…

AID, a novel training-free method, significantly improves image interpolation by fusing inner/outer interpolated attention layers and using beta-distribution for coefficient selection, enhancing consi…

Accelerate Schrödinger Bridge Matching with Discrete-time IMF using only a few steps, achieving comparable results to existing hundred-step methods via D-GAN implementation.

FG-DMs revolutionize image synthesis by jointly modeling image and condition distributions, achieving higher object recall and enabling flexible editing.

TwinAct: Decoupling actions and actors for customizable text-guided action image generation.

ACFun: A novel facial stylization method fusing abstract & concrete features for high-quality, artistically pleasing results from only one style & one face image.

A novel modular diffusion framework for image reconstruction dramatically cuts computational costs and achieves state-of-the-art perceptual quality across various tasks by cleverly combining pre-train…

2DQuant achieves highly efficient and accurate low-bit image super-resolution by using a dual-stage post-training quantization method that minimizes accuracy loss in transformer-based models, surpassi…