Diffusion-based Methods. With the recent explosion of interest in DMs, many have attempted to employ them for unpaired I2I translation. For instance, ILVR [6] achieves an unpaired image translation by modifying the standard Gaussian denoising process. It relies on a DM trained only on the targe...
To refine the image in the latent space, we propose a latent time-travel strategy that increases the sampling timestep. Empirically, we compare Latent-SDE to the baseline of the score-based diffusion model on three widely adopted unpaired I2I tasks under two metrics. Latent-SDE achieves state-...
We propose a novel unpaired image-to-image translation method that uses denoising diffusion probabilistic models without requiring adversarial training. Our method, UNpaired Image Translation with Denoising Diffusion Probabilistic Models (UNIT-DDPM), trains a generative model to infer the joint distribution...
Full size image Non-reciprocal reflections Finally, we show that the unpaired Dirac point enables the phenomenon of non-reciprocal reflection. As shown in Fig.4a, a diverging beam is incident on the upper boundary of the PhC slab (θ = 12.9° andB = 0.4 Tesla) from the left-upp...
One-step image-to-image with Stable Diffusion turbo: sketch2image, day2night, and more - unpaired training · GaParmar/img2img-turbo@db6cf92
potentially compromising theirperformance and safety. To address this, our paper in-troduces LightDiff, a domain-tailored framework designedto enhance the low-light image quality for autonomousdriving applications. Specif i cally, we employ a multi-condition controlled diffusion model. LightDiff works ...
the unique challenges, including warping garments to more diverse human poses and rendering more complex backgrounds faithfully, by a novel DensePose warping correction method combined with diffusion-based conditional inpainting. Our experiments show competitive performance for standard studio try-on tasks ...
Recent score-based diffusion models (SBDMs) show promising results in unpaired image-to-image translation (I2I). However, existing methods, either energy-based or statistically-based, provide no explicit form of the interfered intermediate generative distributions. This work presents a new score-decomp...
In this paper, we propose a target-guided diffusion model (TGDM) for unpaired cross-modality medical image translation. For training, to encourage our diffusion model to learn more visual concepts, we adopted a perception prioritized weight scheme (P2W) to the training objectives. For sampling, ...
This study introduces a novel structure-preserving diffusion model specifically designed for unpaired medical image translation, leveraging edge information to represent common anatomical structures across different modalities. To bridge the domain gap effectively, we further propose a novel Interleaved Sampling...