What's the score?

Existing score distillation methods are sensitive to classifier-free guidance (CFG) scale: manifested as over-smoothness or instability at small CFG scales, while over-saturation at large ones. To explain and analyze these issues, we revisit the derivation of Score Distillation Sampling (SDS) and decipher existing score distillation with the Wasserstein Generative Adversarial Network (WGAN) paradigm. With the WGAN paradigm, we find that existing score distillation either employs a fixed sub-optimal discriminator or conducts incomplete discriminator optimization, resulting in the scale-sensitive issue. We propose the Adversarial Score Distillation (ASD), which maintains an optimizable discriminator and updates it using the complete optimization objective. Experiments show that the proposed ASD performs favorably in 2D distillation and text-to-3D tasks against existing methods. Furthermore, to explore the generalization ability of our WGAN paradigm, we extend ASD to the image editing task, which achieves competitive results. The project page and code are at https://github.com/2y7c3/ASD.

Diffusion models have gained prominence in generating data for perception tasks such as image classification and object detection. However, the potential in generating high-quality tracking sequences, a crucial aspect in the field of video perception, has not been fully investigated. To address this gap, we propose TrackDiffusion, a novel architecture designed to generate continuous video sequences from the tracklets. TrackDiffusion represents a significant departure from the traditional layout-to-image (L2I) generation and copy-paste synthesis focusing on static image elements like bounding boxes by empowering image diffusion models to encompass dynamic and continuous tracking trajectories, thereby capturing complex motion nuances and ensuring instance consistency among video frames. For the first time, we demonstrate that the generated video sequences can be utilized for training multi-object tracking (MOT) systems, leading to significant improvement in tracker performance. Experimental results show that our model significantly enhances instance consistency in generated video sequences, leading to improved perceptual metrics. Our approach achieves an improvement of 8.7 in TrackAP and 11.8 in TrackAP$_{50}$ on the YTVIS dataset, underscoring its potential to redefine the standards of video data generation for MOT tasks and beyond.

We present DREAM, a novel training framework representing Diffusion Rectification and Estimation-Adaptive Models, requiring minimal code changes (just three lines) yet significantly enhancing the alignment of training with sampling in diffusion models. DREAM features two components: diffusion rectification, which adjusts training to reflect the sampling process, and estimation adaptation, which balances perception against distortion. When applied to image super-resolution (SR), DREAM adeptly navigates the tradeoff between minimizing distortion and preserving high image quality. Experiments demonstrate DREAM’s superiority over standard diffusion-based SR methods, showing a $2$ to $3\times $ faster training convergence and a $10$ to $20\times$ reduction in necessary sampling steps to achieve comparable or superior results. We hope DREAM will inspire a rethinking of diffusion model training paradigms.

Image-to-image translation (I2IT) refers to the process of transforming images from a source domain to a target domain while maintaining a fundamental connection in terms of image content. In the past few years, remarkable advancements in I2IT were achieved by Generative Adversarial Networks (GANs), which nevertheless struggle with translations requiring high precision. Recently, Diffusion Models have established themselves as the engine of choice for image generation. In this paper we introduce S2ST, a novel framework designed to accomplish global I2IT in complex photorealistic images, such as day-to-night or clear-to-rain translations of automotive scenes. S2ST operates within the seed space of a Latent Diffusion Model, thereby leveraging the powerful image priors learned by the latter. We show that S2ST surpasses state-of-the-art GAN-based I2IT methods, as well as diffusion-based approaches, for complex automotive scenes, improving fidelity while respecting the target domain’s appearance across a variety of domains. Notably, S2ST obviates the necessity for training domain-specific translation networks.

Contents generated by recent advanced Text-to-Image (T2I) diffusion models are sometimes too imaginative for existing off-the-shelf property semantic predictors to estimate due to the immitigable domain gap. We introduce DMP, a pipeline utilizing pre-trained T2I models as a prior for pixel-level semantic prediction tasks. To address the misalignment between deterministic prediction tasks and stochastic T2I models, we reformulate the diffusion process through a sequence of interpolations, establishing a deterministic mapping between input RGB images and output prediction distributions. To preserve generalizability, we use low-rank adaptation to fine-tune pre-trained models. Extensive experiments across five tasks, including 3D property estimation, semantic segmentation, and intrinsic image decomposition, showcase the efficacy of the proposed method. Despite limited-domain training data, the approach yields faithful estimations for arbitrary images, surpassing existing state-of-the-art algorithms.

Diffusion models generate high-quality images but require dozens of forward passes. We introduce Distribution Matching Distillation (DMD), a procedure to transform a diffusion model into a one-step image generator with minimal impact on image quality. We enforce the one-step image generator match the diffusion model at distribution level, by minimizing an approximate KL divergence whose gradient can be expressed as the difference between 2 score functions, one of the target distribution and the other of the synthetic distribution being produced by our one-step generator. The score functions are parameterized as two diffusion models trained separately on each distribution. Combined with a simple regression loss matching the large-scale structure of the multi-step diffusion outputs, our method outperforms all published few-step diffusion approaches, reaching 2.62 FID on ImageNet 64x64 and 11.49 FID on zero-shot COCO-30k, comparable to Stable Diffusion but orders of magnitude faster. Utilizing FP16 inference, our model can generate images at 20 FPS on modern hardware.

Recent work has demonstrated a remarkable ability to customize text-to-image diffusion models to multiple, fine-grained concepts in a sequential (i.e., continual) manner while only providing a few example images for each concept. This setting is known as continual diffusion. Here, we ask the question: Can we scale these methods to longer concept sequences without forgetting? Although prior work mitigates the forgetting of previously learned concepts, we show that its capacity to learn new tasks reaches saturation over longer sequences. We address this challenge by introducing a novel method, STack-And-Mask INcremental Adapters (STAMINA), which is composed of low-ranked attention-masked adapters and customized MLP tokens. STAMINA is designed to enhance the robust fine-tuning properties of LoRA for sequential concept learning via learnable hard-attention masks parameterized with low rank MLPs, enabling precise, scalable learning via sparse adaptation. Notably, all introduced trainable parameters can be folded back into the model after training, inducing no additional inference parameter costs. We show that STAMINA outperforms the prior SOTA for the setting of text-to-image continual customization on a 50-concept benchmark composed of landmarks and human faces, with no stored replay data. Additionally, we extended our method to the setting of continual learning for image classification, demonstrating that our gains also translate to state-of-the-art performance in this standard benchmark.

DiffusionAvatars synthesizes a high-fidelity 3D head avatar of a person, offering intuitive control over both pose and expression. We propose a diffusion-based neural renderer that leverages generic 2D priors to produce compelling images of faces. For coarse guidance of the expression and head pose, we render a neural parametric head model (NPHM) from the target viewpoint, which acts as a proxy geometry of the person. Additionally, to enhance the modeling of intricate facial expressions, we condition DiffusionAvatars directly on the expression codes obtained from NPHM via cross-attention. Finally, to synthesize consistent surface details across different viewpoints and expressions, we rig learnable spatial features to the head’s surface via TriPlane lookup in NPHM’s canonical space. We train DiffusionAvatars on RGB videos and corresponding tracked NPHM meshes of a person and test the obtained avatars in both self-reenactment and animation scenarios. Our experiments demonstrate that DiffusionAvatars generates temporally consistent and visually appealing videos for novel poses and expressions of a person, outperforming existing approaches.

Sampling from diffusion models can be treated as solving the corresponding ordinary differential equations (ODEs), with the aim of obtaining an accurate solution with as few number of function evaluations (NFE) as possible. Recently, various fast samplers utilizing higher-order ODE solvers have emerged and achieved better performance than the initial first-order one. However, these numerical methods inherently result in certain approximation errors, which significantly degrades sample quality with extremely small NFE (e.g., around 5). In contrast, based on the geometric observation that each sampling trajectory almost lies in a two-dimensional subspace embedded in the ambient space, we propose Approximate MEan-Direction Solver (AMED-Solver) that eliminates truncation errors by directly learning the mean direction for fast diffusion sampling. Besides, our method can be easily used as a plugin to further improve existing ODE-based samplers. Extensive experiments on image synthesis with the resolution ranging from 32 to 256 demonstrate the effectiveness of our method. With only 5 NFE, we achieve 7.14 FID on CIFAR-10, 13.75 FID on ImageNet 64$\times$64, and 12.79 FID on LSUN Bedroom. Our code is available at https://github.com/zhyzhouu/amed-solver.

Diffusion probabilistic models (DPMs) are a key component in modern generative models. DPM-solvers have achieved reduced latency and enhanced quality significantly, but have posed challenges to find the exact inverse (i.e., finding the initial noise from the given image). Here we investigate the exact inversions for DPM-solvers and propose algorithms to perform them when samples are generated by the first-order as well as higher-order DPM-solvers. For each explicit denoising step in DPM-solvers, we formulated the inversions using implicit methods such as gradient descent or forward step method to ensure the robustness to large classifier-free guidance unlike the prior approach using fixed-point iteration. Experimental results demonstrated that our proposed exact inversion methods significantly reduced the error of both image and noise reconstructions, greatly enhanced the ability to distinguish invisible watermarks and well prevented unintended background changes consistently during image editing. Project page: \url{https://smhongok.github.io/inv-dpm.html}.

In recent advancements in high-fidelity image generation, Denoising Diffusion Probabilistic Models (DDPMs) have emerged as a key player. However, their application at high resolutions presents significant computational challenges. Current methods, such as patchifying, expedite processes in UNet and Transformer architectures but at the expense of representational capacity. Addressing this, we introduce the Diffusion State Space Model (DiffuSSM), an architecture that supplants attention mechanisms with a more scalable state space model backbone. This approach effectively handles higher resolutions without resorting to global compression, thus preserving detailed image representation throughout the diffusion process. Our focus on FLOP-efficient architectures in diffusion training marks a significant step forward. Comprehensive evaluations on both ImageNet and LSUN datasets at two resolutions demonstrate that DiffuSSMs are on par or even outperform existing diffusion models with attention modules in FID and Inception Score metrics while significantly reducing total FLOP usage.

Generative adversarial networks (GANs) usually struggle in learning from highly diverse data, whose underlying manifold is complex. In this work, we revisit the mathematical foundations of GANs, and theoretically reveal that the native adversarial loss for GAN training is insufficient to fix the problem of subsets with positive Lebesgue measure of the generated data manifold lying out of the real data manifold. Instead, we find that score matching serves as a valid solution to this issue thanks to its capability of persistently pushing the generated data points towards the real data manifold. We thereby propose to improve the optimization of GANs with score matching regularity (SMaRt). Regarding the empirical evidences, we first design a toy example to show that training GANs by the aid of a ground-truth score function can help reproduce the real data distribution more accurately, and then confirm that our approach can consistently boost the synthesis performance of various state-of-the-art GANs on real-world datasets with pre-trained diffusion models acting as the approximate score function. For instance, when training Aurora on the ImageNet 64x64 dataset, we manage to improve FID from 8.87 to 7.11, on par with the performance of one-step consistency model. The source code will be made public.

This paper explores advancements in high-fidelity personalized image generation through the utilization of pre-trained text-to-image diffusion models. While previous approaches have made significant strides in generating versatile scenes based on text descriptions and a few input images, challenges persist in maintaining the subject fidelity within the generated images. In this work, we introduce an innovative algorithm named HiFi Tuner to enhance the appearance preservation of objects during personalized image generation. Our proposed method employs a parameter-efficient fine-tuning framework, comprising a denoising process and a pivotal inversion process. Key enhancements include the utilization of mask guidance, a novel parameter regularization technique, and the incorporation of step-wise subject representations to elevate the sample fidelity. Additionally, we propose a reference-guided generation approach that leverages the pivotal inversion of a reference image to mitigate unwanted subject variations and artifacts. We further extend our method to a novel image editing task: substituting the subject in an image through textual manipulations. Experimental evaluations conducted on the DreamBooth dataset using the Stable Diffusion model showcase promising results. Fine-tuning solely on textual embeddings improves CLIP-T score by 3.6 points and improves DINO score by 9.6 points over Textual Inversion. When fine-tuning all parameters, HiFi Tuner improves CLIP-T score by 1.2 points and improves DINO score by 1.2 points over DreamBooth, establishing a new state of the art.

We introduce a new framework called DiffGEPCI for cross-modality generation in magnetic resonance imaging (MRI) using a 2.5D conditional diffusion model. DiffGEPCI can synthesize high-quality Fluid Attenuated Inversion Recovery (FLAIR) and Magnetization Prepared-Rapid Gradient Echo (MPRAGE) images, without acquiring corresponding measurements, by leveraging multi-Gradient-Recalled Echo (mGRE) MRI signals as conditional inputs. DiffGEPCI operates in a two-step fashion: it initially estimates a 3D volume slice-by-slice using the axial plane and subsequently applies a refinement algorithm (referred to as 2.5D) to enhance the quality of the coronal and sagittal planes. Experimental validation on real mGRE data shows that DiffGEPCI achieves excellent performance, surpassing generative adversarial networks (GANs) and traditional diffusion models.

While many unsupervised learning models focus on one family of tasks, either generative or discriminative, we explore the possibility of a unified representation learner: a model which addresses both families of tasks simultaneously. We identify diffusion models, a state-of-the-art method for generative tasks, as a prime candidate. Such models involve training a U-Net to iteratively predict and remove noise, and the resulting model can synthesize high-fidelity, diverse, novel images. We find that the intermediate feature maps of the U-Net are diverse, discriminative feature representations. We propose a novel attention mechanism for pooling feature maps and further leverage this mechanism as DifFormer, a transformer feature fusion of features from different diffusion U-Net blocks and noise steps. We also develop DifFeed, a novel feedback mechanism tailored to diffusion. We find that diffusion models are better than GANs, and, with our fusion and feedback mechanisms, can compete with state-of-the-art unsupervised image representation learning methods for discriminative tasks - image classification with full and semi-supervision, transfer for fine-grained classification, object detection and segmentation, and semantic segmentation. Our project website (https://mgwillia.github.io/diffssl/) and code (https://github.com/soumik-kanad/diffssl) are available publicly.

We introduce SODA, a self-supervised diffusion model, designed for representation learning. The model incorporates an image encoder, which distills a source view into a compact representation, that, in turn, guides the generation of related novel views. We show that by imposing a tight bottleneck between the encoder and a denoising decoder, and leveraging novel view synthesis as a self-supervised objective, we can turn diffusion models into strong representation learners, capable of capturing visual semantics in an unsupervised manner. To the best of our knowledge, SODA is the first diffusion model to succeed at ImageNet linear-probe classification, and, at the same time, it accomplishes reconstruction, editing and synthesis tasks across a wide range of datasets. Further investigation reveals the disentangled nature of its emergent latent space, that serves as an effective interface to control and manipulate the model’s produced images. All in all, we aim to shed light on the exciting and promising potential of diffusion models, not only for image generation, but also for learning rich and robust representations.

Most real-world networks are noisy and incomplete samples from an unknown target distribution. Refining them by correcting corruptions or inferring unobserved regions typically improves downstream performance. Inspired by the impressive generative capabilities that have been used to correct corruptions in images, and the similarities between “in-painting” and filling in missing nodes and edges conditioned on the observed graph, we propose a novel graph generative framework, SGDM, which is based on subgraph diffusion. Our framework not only improves the scalability and fidelity of graph diffusion models, but also leverages the reverse process to perform novel, conditional generation tasks. In particular, through extensive empirical analysis and a set of novel metrics, we demonstrate that our proposed model effectively supports the following refinement tasks for partially observable networks: T1: denoising extraneous subgraphs, T2: expanding existing subgraphs and T3: performing “style” transfer by regenerating a particular subgraph to match the characteristics of a different node or subgraph.

In this paper, we address the limitations of existing text-to-image diffusion models in generating demographically fair results when given human-related descriptions. These models often struggle to disentangle the target language context from sociocultural biases, resulting in biased image generation. To overcome this challenge, we propose Fair Mapping, a general, model-agnostic, and lightweight approach that modifies a pre-trained text-to-image model by controlling the prompt to achieve fair image generation. One key advantage of our approach is its high efficiency. The training process only requires updating a small number of parameters in an additional linear mapping network. This not only reduces the computational cost but also accelerates the optimization process. We first demonstrate the issue of bias in generated results caused by language biases in text-guided diffusion models. By developing a mapping network that projects language embeddings into an unbiased space, we enable the generation of relatively balanced demographic results based on a keyword specified in the prompt. With comprehensive experiments on face image generation, we show that our method significantly improves image generation performance when prompted with descriptions related to human faces. By effectively addressing the issue of bias, we produce more fair and diverse image outputs. This work contributes to the field of text-to-image generation by enhancing the ability to generate images that accurately reflect the intended demographic characteristics specified in the text.

The aim of this short note is to show that Denoising Diffusion Probabilistic Model DDPM, a non-homogeneous discrete-time Markov process, can be represented by a time-homogeneous continuous-time Markov process observed at non-uniformly sampled discrete times. Surprisingly, this continuous-time Markov process is the well-known and well-studied Ornstein-Ohlenbeck (OU) process, which was developed in 1930’s for studying Brownian particles in Harmonic potentials. We establish the formal equivalence between DDPM and the OU process using its analytical solution. We further demonstrate that the design problem of the noise scheduler for non-homogeneous DDPM is equivalent to designing observation times for the OU process. We present several heuristic designs for observation times based on principled quantities such as auto-variance and Fisher Information and connect them to ad hoc noise schedules for DDPM. Interestingly, we show that the Fisher-Information-motivated schedule corresponds exactly the cosine schedule, which was developed without any theoretical foundation but is the current state-of-the-art noise schedule.

The booming use of text-to-image generative models has raised concerns about their high risk of producing copyright-infringing content. While probabilistic copyright protection methods provide a probabilistic guarantee against such infringement, in this paper, we introduce Virtually Assured Amplification Attack (VA3), a novel online attack framework that exposes the vulnerabilities of these protection mechanisms. The proposed framework significantly amplifies the probability of generating infringing content on the sustained interactions with generative models and a lower-bounded success probability of each engagement. Our theoretical and experimental results demonstrate the effectiveness of our approach and highlight the potential risk of implementing probabilistic copyright protection in practical applications of text-to-image generative models. Code is available at https://github.com/South7X/VA3.

We introduce HiDiffusion, a tuning-free framework comprised of Resolution-Aware U-Net (RAU-Net) and Modified Shifted Window Multi-head Self-Attention (MSW-MSA) to enable pretrained large text-to-image diffusion models to efficiently generate high-resolution images (e.g. 1024$\times$1024) that surpass the training image resolution. Pretrained diffusion models encounter unreasonable object duplication in generating images beyond the training image resolution. We attribute it to the mismatch between the feature map size of high-resolution images and the receptive field of U-Net’s convolution. To address this issue, we propose a simple yet scalable method named RAU-Net. RAU-Net dynamically adjusts the feature map size to match the convolution’s receptive field in the deep block of U-Net. Another obstacle in high-resolution synthesis is the slow inference speed of U-Net. Our observations reveal that the global self-attention in the top block, which exhibits locality, however, consumes the majority of computational resources. To tackle this issue, we propose MSW-MSA. Unlike previous window attention mechanisms, our method uses a much larger window size and dynamically shifts windows to better accommodate diffusion models. Extensive experiments demonstrate that our HiDiffusion can scale diffusion models to generate 1024$\times$1024, 2048$\times$2048, or even 4096$\times$4096 resolution images, while simultaneously reducing inference time by 40\%-60\%, achieving state-of-the-art performance on high-resolution image synthesis. The most significant revelation of our work is that a pretrained diffusion model on low-resolution images is scalable for high-resolution generation without further tuning. We hope this revelation can provide insights for future research on the scalability of diffusion models.

In recent years, Text-to-Image (T2I) models have seen remarkable advancements, gaining widespread adoption. However, this progress has inadvertently opened avenues for potential misuse, particularly in generating inappropriate or Not-Safe-For-Work (NSFW) content. Our work introduces MMA-Diffusion, a framework that presents a significant and realistic threat to the security of T2I models by effectively circumventing current defensive measures in both open-source models and commercial online services. Unlike previous approaches, MMA-Diffusion leverages both textual and visual modalities to bypass safeguards like prompt filters and post-hoc safety checkers, thus exposing and highlighting the vulnerabilities in existing defense mechanisms.

Text-to-image diffusion models have remarkably excelled in producing diverse, high-quality, and photo-realistic images. This advancement has spurred a growing interest in incorporating specific identities into generated content. Most current methods employ an inversion approach to embed a target visual concept into the text embedding space using a single reference image. However, the newly synthesized faces either closely resemble the reference image in terms of facial attributes, such as expression, or exhibit a reduced capacity for identity preservation. Text descriptions intended to guide the facial attributes of the synthesized face may fall short, owing to the intricate entanglement of identity information with identity-irrelevant facial attributes derived from the reference image. To address these issues, we present the novel use of the extended StyleGAN embedding space $\mathcal{W}_+$, to achieve enhanced identity preservation and disentanglement for diffusion models. By aligning this semantically meaningful human face latent space with text-to-image diffusion models, we succeed in maintaining high fidelity in identity preservation, coupled with the capacity for semantic editing. Additionally, we propose new training objectives to balance the influences of both prompt and identity conditions, ensuring that the identity-irrelevant background remains unaffected during facial attribute modifications. Extensive experiments reveal that our method adeptly generates personalized text-to-image outputs that are not only compatible with prompt descriptions but also amenable to common StyleGAN editing directions in diverse settings. Our source code will be available at \url{https://github.com/csxmli2016/w-plus-adapter}.

Scene flow estimation, which aims to predict per-point 3D displacements of dynamic scenes, is a fundamental task in the computer vision field. However, previous works commonly suffer from unreliable correlation caused by locally constrained searching ranges, and struggle with accumulated inaccuracy arising from the coarse-to-fine structure. To alleviate these problems, we propose a novel uncertainty-aware scene flow estimation network (DifFlow3D) with the diffusion probabilistic model. Iterative diffusion-based refinement is designed to enhance the correlation robustness and resilience to challenging cases, e.g., dynamics, noisy inputs, repetitive patterns, etc. To restrain the generation diversity, three key flow-related features are leveraged as conditions in our diffusion model. Furthermore, we also develop an uncertainty estimation module within diffusion to evaluate the reliability of estimated scene flow. Our DifFlow3D achieves state-of-the-art performance, with 6.7\% and 19.1\% EPE3D reduction respectively on FlyingThings3D and KITTI 2015 datasets. Notably, our method achieves an unprecedented millimeter-level accuracy (0.0089m in EPE3D) on the KITTI dataset. Additionally, our diffusion-based refinement paradigm can be readily integrated as a plug-and-play module into existing scene flow networks, significantly increasing their estimation accuracy. Codes will be released later.

We propose CompFuser, an image generation pipeline that enhances spatial comprehension and attribute assignment in text-to-image generative models. Our pipeline enables the interpretation of instructions defining spatial relationships between objects in a scene, such as `An image of a gray cat on the left of an orange dog’, and generate corresponding images. This is especially important in order to provide more control to the user. CompFuser overcomes the limitation of existing text-to-image diffusion models by decoding the generation of multiple objects into iterative steps: first generating a single object and then editing the image by placing additional objects in their designated positions. To create training data for spatial comprehension and attribute assignment we introduce a synthetic data generation process, that leverages a frozen large language model and a frozen layout-based diffusion model for object placement. We compare our approach to strong baselines and show that our model outperforms state-of-the-art image generation models in spatial comprehension and attribute assignment, despite being 3x to 5x smaller in parameters.

Generative models can produce impressively realistic images. This paper demonstrates that generated images have geometric features different from those of real images. We build a set of collections of generated images, prequalified to fool simple, signal-based classifiers into believing they are real. We then show that prequalified generated images can be identified reliably by classifiers that only look at geometric properties. We use three such classifiers. All three classifiers are denied access to image pixels, and look only at derived geometric features. The first classifier looks at the perspective field of the image, the second looks at lines detected in the image, and the third looks at relations between detected objects and shadows. Our procedure detects generated images more reliably than SOTA local signal based detectors, for images from a number of distinct generators. Saliency maps suggest that the classifiers can identify geometric problems reliably. We conclude that current generators cannot reliably reproduce geometric properties of real images.

Nature evolves creatures with a high complexity of morphological and behavioral intelligence, meanwhile computational methods lag in approaching that diversity and efficacy. Co-optimization of artificial creatures’ morphology and control in silico shows promise for applications in physical soft robotics and virtual character creation; such approaches, however, require developing new learning algorithms that can reason about function atop pure structure. In this paper, we present DiffuseBot, a physics-augmented diffusion model that generates soft robot morphologies capable of excelling in a wide spectrum of tasks. DiffuseBot bridges the gap between virtually generated content and physical utility by (i) augmenting the diffusion process with a physical dynamical simulation which provides a certificate of performance, and (ii) introducing a co-design procedure that jointly optimizes physical design and control by leveraging information about physical sensitivities from differentiable simulation. We showcase a range of simulated and fabricated robots along with their capabilities. Check our website at https://diffusebot.github.io/

We introduce Adversarial Diffusion Distillation (ADD), a novel training approach that efficiently samples large-scale foundational image diffusion models in just 1-4 steps while maintaining high image quality. We use score distillation to leverage large-scale off-the-shelf image diffusion models as a teacher signal in combination with an adversarial loss to ensure high image fidelity even in the low-step regime of one or two sampling steps. Our analyses show that our model clearly outperforms existing few-step methods (GANs, Latent Consistency Models) in a single step and reaches the performance of state-of-the-art diffusion models (SDXL) in only four steps. ADD is the first method to unlock single-step, real-time image synthesis with foundation models. Code and weights available under https://github.com/Stability-AI/generative-models and https://huggingface.co/stabilityai/ .

We present a new method for text-driven motion transfer - synthesizing a video that complies with an input text prompt describing the target objects and scene while maintaining an input video’s motion and scene layout. Prior methods are confined to transferring motion across two subjects within the same or closely related object categories and are applicable for limited domains (e.g., humans). In this work, we consider a significantly more challenging setting in which the target and source objects differ drastically in shape and fine-grained motion characteristics (e.g., translating a jumping dog into a dolphin). To this end, we leverage a pre-trained and fixed text-to-video diffusion model, which provides us with generative and motion priors. The pillar of our method is a new space-time feature loss derived directly from the model. This loss guides the generation process to preserve the overall motion of the input video while complying with the target object in terms of shape and fine-grained motion traits.

Existing text-to-image (T2I) diffusion models usually struggle in interpreting complex prompts, especially those with quantity, object-attribute binding, and multi-subject descriptions. In this work, we introduce a semantic panel as the middleware in decoding texts to images, supporting the generator to better follow instructions. The panel is obtained through arranging the visual concepts parsed from the input text by the aid of large language models, and then injected into the denoising network as a detailed control signal to complement the text condition. To facilitate text-to-panel learning, we come up with a carefully designed semantic formatting protocol, accompanied by a fully-automatic data preparation pipeline. Thanks to such a design, our approach, which we call Ranni, manages to enhance a pre-trained T2I generator regarding its textual controllability. More importantly, the introduction of the generative middleware brings a more convenient form of interaction (i.e., directly adjusting the elements in the panel or using language instructions) and further allows users to finely customize their generation, based on which we develop a practical system and showcase its potential in continuous generation and chatting-based editing. Our project page is at https://ranni-t2i.github.io/Ranni.

Underwater images are subject to intricate and diverse degradation, inevitably affecting the effectiveness of underwater visual tasks. However, most approaches primarily operate in the raw pixel space of images, which limits the exploration of the frequency characteristics of underwater images, leading to an inadequate utilization of deep models’ representational capabilities in producing high-quality images. In this paper, we introduce a novel Underwater Image Enhancement (UIE) framework, named WF-Diff, designed to fully leverage the characteristics of frequency domain information and diffusion models. WF-Diff consists of two detachable networks: Wavelet-based Fourier information interaction network (WFI2-net) and Frequency Residual Diffusion Adjustment Module (FRDAM). With our full exploration of the frequency domain information, WFI2-net aims to achieve preliminary enhancement of frequency information in the wavelet space. Our proposed FRDAM can further refine the high- and low-frequency information of the initial enhanced images, which can be viewed as a plug-and-play universal module to adjust the detail of the underwater images. With the above techniques, our algorithm can show SOTA performance on real-world underwater image datasets, and achieves competitive performance in visual quality.

Recent advances in imaging and high-performance computing have made it possible to image the entire human brain at the cellular level. This is the basis to study the multi-scale architecture of the brain regarding its subdivision into brain areas and nuclei, cortical layers, columns, and cell clusters down to single cell morphology Methods for brain mapping and cell segmentation exploit such images to enable rapid and automated analysis of cytoarchitecture and cell distribution in complete series of histological sections. However, the presence of inevitable processing artifacts in the image data caused by missing sections, tears in the tissue, or staining variations remains the primary reason for gaps in the resulting image data. To this end we aim to provide a model that can fill in missing information in a reliable way, following the true cell distribution at different scales. Inspired by the recent success in image generation, we propose a denoising diffusion probabilistic model (DDPM), trained on light-microscopic scans of cell-body stained sections. We extend this model with the RePaint method to impute missing or replace corrupted image data. We show that our trained DDPM is able to generate highly realistic image information for this purpose, generating plausible cell statistics and cytoarchitectonic patterns. We validate its outputs using two established downstream task models trained on the same data.

Diffusion models, a type of generative model, have demonstrated great potential for synthesizing highly detailed images. By integrating with GAN, advanced diffusion models like DDGAN \citep{xiao2022DDGAN} could approach real-time performance for expansive practical applications. While DDGAN has effectively addressed the challenges of generative modeling, namely producing high-quality samples, covering different data modes, and achieving faster sampling, it remains susceptible to performance drops caused by datasets that are corrupted with outlier samples. This work introduces a robust training technique based on semi-unbalanced optimal transport to mitigate the impact of outliers effectively. Through comprehensive evaluations, we demonstrate that our robust diffusion GAN (RDGAN) outperforms vanilla DDGAN in terms of the aforementioned generative modeling criteria, i.e., image quality, mode coverage of distribution, and inference speed, and exhibits improved robustness when dealing with both clean and corrupted datasets.

The deployment of large-scale text-to-image diffusion models on mobile devices is impeded by their substantial model size and slow inference speed. In this paper, we propose \textbf{MobileDiffusion}, a highly efficient text-to-image diffusion model obtained through extensive optimizations in both architecture and sampling techniques. We conduct a comprehensive examination of model architecture design to reduce redundancy, enhance computational efficiency, and minimize model’s parameter count, while preserving image generation quality. Additionally, we employ distillation and diffusion-GAN finetuning techniques on MobileDiffusion to achieve 8-step and 1-step inference respectively. Empirical studies, conducted both quantitatively and qualitatively, demonstrate the effectiveness of our proposed techniques. MobileDiffusion achieves a remarkable \textbf{sub-second} inference speed for generating a $512\times512$ image on mobile devices, establishing a new state of the art.

Speech-driven 3D facial animation has been an attractive task in both academia and industry. Traditional methods mostly focus on learning a deterministic mapping from speech to animation. Recent approaches start to consider the non-deterministic fact of speech-driven 3D face animation and employ the diffusion model for the task. However, personalizing facial animation and accelerating animation generation are still two major limitations of existing diffusion-based methods. To address the above limitations, we propose DiffusionTalker, a diffusion-based method that utilizes contrastive learning to personalize 3D facial animation and knowledge distillation to accelerate 3D animation generation. Specifically, to enable personalization, we introduce a learnable talking identity to aggregate knowledge in audio sequences. The proposed identity embeddings extract customized facial cues across different people in a contrastive learning manner. During inference, users can obtain personalized facial animation based on input audio, reflecting a specific talking style. With a trained diffusion model with hundreds of steps, we distill it into a lightweight model with 8 steps for acceleration. Extensive experiments are conducted to demonstrate that our method outperforms state-of-the-art methods. The code will be released.

Flow matching as a paradigm of generative model achieves notable success across various domains. However, existing methods use either multi-round training or knowledge within minibatches, posing challenges in finding a favorable coupling strategy for straight trajectories. To address this issue, we propose a novel approach, Straighter trajectories of Flow Matching (StraightFM). It straightens trajectories with the coupling strategy guided by diffusion model from entire distribution level. First, we propose a coupling strategy to straighten trajectories, creating couplings between image and noise samples under diffusion model guidance. Second, StraightFM also integrates real data to enhance training, employing a neural network to parameterize another coupling process from images to noise samples. StraightFM is jointly optimized with couplings from above two mutually complementary directions, resulting in straighter trajectories and enabling both one-step and few-step generation. Extensive experiments demonstrate that StraightFM yields high quality samples with fewer step. StraightFM generates visually appealing images with a lower FID among diffusion and traditional flow matching methods within 5 sampling steps when trained on pixel space. In the latent space (i.e., Latent Diffusion), StraightFM achieves a lower KID value compared to existing methods on the CelebA-HQ 256 dataset in fewer than 10 sampling steps.

Diffusion models have shown great potential for vision-related tasks, particularly for image generation. However, their training is typically conducted in a centralized manner, relying on data collected from publicly available sources. This approach may not be feasible or practical in many domains, such as the medical field, which involves privacy concerns over data collection. Despite the challenges associated with privacy-sensitive data, such domains could still benefit from valuable vision services provided by diffusion models. Federated learning (FL) plays a crucial role in enabling decentralized model training without compromising data privacy. Instead of collecting data, an FL system gathers model parameters, effectively safeguarding the private data of different parties involved. This makes FL systems vital for managing decentralized learning tasks, especially in scenarios where privacy-sensitive data is distributed across a network of clients. Nonetheless, FL presents its own set of challenges due to its distributed nature and privacy-preserving properties. Therefore, in this study, we explore the FL strategy to train diffusion models, paving the way for the development of federated diffusion models. We conduct experiments on various FL scenarios, and our findings demonstrate that federated diffusion models have great potential to deliver vision services to privacy-sensitive domains.

The diffusion model has been proven a powerful generative model in recent years, yet remains a challenge in generating visual text. Several methods alleviated this issue by incorporating explicit text position and content as guidance on where and what text to render. However, these methods still suffer from several drawbacks, such as limited flexibility and automation, constrained capability of layout prediction, and restricted style diversity. In this paper, we present TextDiffuser-2, aiming to unleash the power of language models for text rendering. Firstly, we fine-tune a large language model for layout planning. The large language model is capable of automatically generating keywords for text rendering and also supports layout modification through chatting. Secondly, we utilize the language model within the diffusion model to encode the position and texts at the line level. Unlike previous methods that employed tight character-level guidance, this approach generates more diverse text images. We conduct extensive experiments and incorporate user studies involving human participants as well as GPT-4V, validating TextDiffuser-2’s capacity to achieve a more rational text layout and generation with enhanced diversity. The code and model will be available at \url{https://aka.ms/textdiffuser-2}.

Text-to-image diffusion models are well-known for their ability to generate realistic images based on textual prompts. However, the existing works have predominantly focused on English, lacking support for non-English text-to-image models. The most commonly used translation methods cannot solve the generation problem related to language culture, while training from scratch on a specific language dataset is prohibitively expensive. In this paper, we are inspired to propose a simple plug-and-play language transfer method based on knowledge distillation. All we need to do is train a lightweight MLP-like parameter-efficient adapter (PEA) with only 6M parameters under teacher knowledge distillation along with a small parallel data corpus. We are surprised to find that freezing the parameters of UNet can still achieve remarkable performance on the language-specific prompt evaluation set, demonstrating that PEA can stimulate the potential generation ability of the original UNet. Additionally, it closely approaches the performance of the English text-to-image model on a general prompt evaluation set. Furthermore, our adapter can be used as a plugin to achieve significant results in downstream tasks in cross-lingual text-to-image generation. Code will be available at: https://github.com/OPPO-Mente-Lab/PEA-Diffusion

Despite the recent advancements, conditional image generation still faces challenges of cost, generalizability, and the need for task-specific training. In this paper, we propose Manifold Preserving Guided Diffusion (MPGD), a training-free conditional generation framework that leverages pretrained diffusion models and off-the-shelf neural networks with minimal additional inference cost for a broad range of tasks. Specifically, we leverage the manifold hypothesis to refine the guided diffusion steps and introduce a shortcut algorithm in the process. We then propose two methods for on-manifold training-free guidance using pre-trained autoencoders and demonstrate that our shortcut inherently preserves the manifolds when applied to latent diffusion models. Our experiments show that MPGD is efficient and effective for solving a variety of conditional generation applications in low-compute settings, and can consistently offer up to 3.8x speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines.

The advancements in generative modeling, particularly the advent of diffusion models, have sparked a fundamental question: how can these models be effectively used for discriminative tasks? In this work, we find that generative models can be great test-time adapters for discriminative models. Our method, Diffusion-TTA, adapts pre-trained discriminative models such as image classifiers, segmenters and depth predictors, to each unlabelled example in the test set using generative feedback from a diffusion model. We achieve this by modulating the conditioning of the diffusion model using the output of the discriminative model. We then maximize the image likelihood objective by backpropagating the gradients to discriminative model’s parameters. We show Diffusion-TTA significantly enhances the accuracy of various large-scale pre-trained discriminative models, such as, ImageNet classifiers, CLIP models, image pixel labellers and image depth predictors. Diffusion-TTA outperforms existing test-time adaptation methods, including TTT-MAE and TENT, and particularly shines in online adaptation setups, where the discriminative model is continually adapted to each example in the test set. We provide access to code, results, and visualizations on our website: https://diffusion-tta.github.io/.

Text-to-image generation has witnessed significant progress with the advent of diffusion models. Despite the ability to generate photorealistic images, current text-to-image diffusion models still often struggle to accurately interpret and follow complex input text prompts. In contrast to existing models that aim to generate images only with their best effort, we introduce Self-correcting LLM-controlled Diffusion (SLD). SLD is a framework that generates an image from the input prompt, assesses its alignment with the prompt, and performs self-corrections on the inaccuracies in the generated image. Steered by an LLM controller, SLD turns text-to-image generation into an iterative closed-loop process, ensuring correctness in the resulting image. SLD is not only training-free but can also be seamlessly integrated with diffusion models behind API access, such as DALL-E 3, to further boost the performance of state-of-the-art diffusion models. Experimental results show that our approach can rectify a majority of incorrect generations, particularly in generative numeracy, attribute binding, and spatial relationships. Furthermore, by simply adjusting the instructions to the LLM, SLD can perform image editing tasks, bridging the gap between text-to-image generation and image editing pipelines. We will make our code available for future research and applications.

Since American Sign Language (ASL) has no standard written form, Deaf signers frequently share videos in order to communicate in their native language. However, since both hands and face convey critical linguistic information in signed languages, sign language videos cannot preserve signer privacy. While signers have expressed interest, for a variety of applications, in sign language video anonymization that would effectively preserve linguistic content, attempts to develop such technology have had limited success, given the complexity of hand movements and facial expressions. Existing approaches rely predominantly on precise pose estimations of the signer in video footage and often require sign language video datasets for training. These requirements prevent them from processing videos ‘in the wild,’ in part because of the limited diversity present in current sign language video datasets. To address these limitations, our research introduces DiffSLVA, a novel methodology that utilizes pre-trained large-scale diffusion models for zero-shot text-guided sign language video anonymization. We incorporate ControlNet, which leverages low-level image features such as HED (Holistically-Nested Edge Detection) edges, to circumvent the need for pose estimation. Additionally, we develop a specialized module dedicated to capturing facial expressions, which are critical for conveying essential linguistic information in signed languages. We then combine the above methods to achieve anonymization that better preserves the essential linguistic content of the original signer. This innovative methodology makes possible, for the first time, sign language video anonymization that could be used for real-world applications, which would offer significant benefits to the Deaf and Hard-of-Hearing communities. We demonstrate the effectiveness of our approach with a series of signer anonymization experiments.

Score-based diffusion models have emerged as one of the most promising frameworks for deep generative modelling, due to their state-of-the art performance in many generation tasks while relying on mathematical foundations such as stochastic differential equations (SDEs) and ordinary differential equations (ODEs). Empirically, it has been reported that ODE based samples are inferior to SDE based samples. In this paper we rigorously describe the range of dynamics and approximations that arise when training score-based diffusion models, including the true SDE dynamics, the neural approximations, the various approximate particle dynamics that result, as well as their associated Fokker–Planck equations and the neural network approximations of these Fokker–Planck equations. We systematically analyse the difference between the ODE and SDE dynamics of score-based diffusion models, and link it to an associated Fokker–Planck equation. We derive a theoretical upper bound on the Wasserstein 2-distance between the ODE- and SDE-induced distributions in terms of a Fokker–Planck residual. We also show numerically that conventional score-based diffusion models can exhibit significant differences between ODE- and SDE-induced distributions which we demonstrate using explicit comparisons. Moreover, we show numerically that reducing the Fokker–Planck residual by adding it as an additional regularisation term leads to closing the gap between ODE- and SDE-induced distributions. Our experiments suggest that this regularisation can improve the distribution generated by the ODE, however that this can come at the cost of degraded SDE sample quality.

Anticipating future actions is inherently uncertain. Given an observed video segment containing ongoing actions, multiple subsequent actions can plausibly follow. This uncertainty becomes even larger when predicting far into the future. However, the majority of existing action anticipation models adhere to a deterministic approach, neglecting to account for future uncertainties. In this work, we rethink action anticipation from a generative view, employing diffusion models to capture different possible future actions. In this framework, future actions are iteratively generated from standard Gaussian noise in the latent space, conditioned on the observed video, and subsequently transitioned into the action space. Extensive experiments on four benchmark datasets, i.e., Breakfast, 50Salads, EpicKitchens, and EGTEA Gaze+, are performed and the proposed method achieves superior or comparable results to state-of-the-art methods, showing the effectiveness of a generative approach for action anticipation. Our code and trained models will be published on GitHub.

The Diffusion model, a prevalent framework for image generation, encounters significant challenges in terms of broad applicability due to its extended inference times and substantial memory requirements. Efficient Post-training Quantization (PTQ) is pivotal for addressing these issues in traditional models. Different from traditional models, diffusion models heavily depend on the time-step $t$ to achieve satisfactory multi-round denoising. Usually, $t$ from the finite set ${1, \ldots, T}$ is encoded to a temporal feature by a few modules totally irrespective of the sampling data. However, existing PTQ methods do not optimize these modules separately. They adopt inappropriate reconstruction targets and complex calibration methods, resulting in a severe disturbance of the temporal feature and denoising trajectory, as well as a low compression efficiency. To solve these, we propose a Temporal Feature Maintenance Quantization (TFMQ) framework building upon a Temporal Information Block which is just related to the time-step $t$ and unrelated to the sampling data. Powered by the pioneering block design, we devise temporal information aware reconstruction (TIAR) and finite set calibration (FSC) to align the full-precision temporal features in a limited time. Equipped with the framework, we can maintain the most temporal information and ensure the end-to-end generation quality. Extensive experiments on various datasets and diffusion models prove our state-of-the-art results. Remarkably, our quantization approach, for the first time, achieves model performance nearly on par with the full-precision model under 4-bit weight quantization. Additionally, our method incurs almost no extra computational cost and accelerates quantization time by $2.0 \times$ on LSUN-Bedrooms $256 \times 256$ compared to previous works.

The recent advent of diffusion models has led to significant progress in solving inverse problems, leveraging these models as effective generative priors. Nonetheless, challenges related to the ill-posed nature of such problems remain, often due to inherent ambiguities in measurements. Drawing inspiration from the human ability to resolve visual ambiguities through perceptual biases, here we introduce a novel latent diffusion inverse solver by incorporating regularization by texts (TReg). Specifically, TReg applies the textual description of the preconception of the solution during the reverse sampling phase, of which description isndynamically reinforced through null-text optimization for adaptive negation. Our comprehensive experimental results demonstrate that TReg successfully mitigates ambiguity in latent diffusion inverse solvers, enhancing their effectiveness and accuracy.

Light field (LF) image super-resolution (SR) is a challenging problem due to its inherent ill-posed nature, where a single low-resolution (LR) input LF image can correspond to multiple potential super-resolved outcomes. Despite this complexity, mainstream LF image SR methods typically adopt a deterministic approach, generating only a single output supervised by pixel-wise loss functions. This tendency often results in blurry and unrealistic results. Although diffusion models can capture the distribution of potential SR results by iteratively predicting Gaussian noise during the denoising process, they are primarily designed for general images and struggle to effectively handle the unique characteristics and information present in LF images. To address these limitations, we introduce LFSRDiff, the first diffusion-based LF image SR model, by incorporating the LF disentanglement mechanism. Our novel contribution includes the introduction of a disentangled U-Net for diffusion models, enabling more effective extraction and fusion of both spatial and angular information within LF images. Through comprehensive experimental evaluations and comparisons with the state-of-the-art LF image SR methods, the proposed approach consistently produces diverse and realistic SR results. It achieves the highest perceptual metric in terms of LPIPS. It also demonstrates the ability to effectively control the trade-off between perception and distortion. The code is available at \url{https://github.com/chaowentao/LFSRDiff}.

We propose a new class of generative diffusion models, called functional diffusion. In contrast to previous work, functional diffusion works on samples that are represented by functions with a continuous domain. Functional diffusion can be seen as an extension of classical diffusion models to an infinite-dimensional domain. Functional diffusion is very versatile as images, videos, audio, 3D shapes, deformations, \etc, can be handled by the same framework with minimal changes. In addition, functional diffusion is especially suited for irregular data or data defined in non-standard domains. In our work, we derive the necessary foundations for functional diffusion and propose a first implementation based on the transformer architecture. We show generative results on complicated signed distance functions and deformation functions defined on 3D surfaces.

Diffusion-based generative models excel in perceptually impressive synthesis but face challenges in interpretability. This paper introduces ToddlerDiffusion, an interpretable 2D diffusion image-synthesis framework inspired by the human generation system. Unlike traditional diffusion models with opaque denoising steps, our approach decomposes the generation process into simpler, interpretable stages; generating contours, a palette, and a detailed colored image. This not only enhances overall performance but also enables robust editing and interaction capabilities. Each stage is meticulously formulated for efficiency and accuracy, surpassing Stable-Diffusion (LDM). Extensive experiments on datasets like LSUN-Churches and COCO validate our approach, consistently outperforming existing methods. ToddlerDiffusion achieves notable efficiency, matching LDM performance on LSUN-Churches while operating three times faster with a 3.76 times smaller architecture. Our source code is provided in the supplementary material and will be publicly accessible.

High-resolution image generation with Generative Artificial Intelligence (GenAI) has immense potential but, due to the enormous capital investment required for training, it is increasingly centralised to a few large corporations, and hidden behind paywalls. This paper aims to democratise high-resolution GenAI by advancing the frontier of high-resolution generation while remaining accessible to a broad audience. We demonstrate that existing Latent Diffusion Models (LDMs) possess untapped potential for higher-resolution image generation. Our novel DemoFusion framework seamlessly extends open-source GenAI models, employing Progressive Upscaling, Skip Residual, and Dilated Sampling mechanisms to achieve higher-resolution image generation. The progressive nature of DemoFusion requires more passes, but the intermediate results can serve as “previews”, facilitating rapid prompt iteration.

We provide full theoretical guarantees for the convergence behaviour of diffusion-based generative models under the assumption of strongly logconcave data distributions while our approximating class of functions used for score estimation is made of Lipschitz continuous functions. We demonstrate via a motivating example, sampling from a Gaussian distribution with unknown mean, the powerfulness of our approach. In this case, explicit estimates are provided for the associated optimization problem, i.e. score approximation, while these are combined with the corresponding sampling estimates. As a result, we obtain the best known upper bound estimates in terms of key quantities of interest, such as the dimension and rates of convergence, for the Wasserstein-2 distance between the data distribution (Gaussian with unknown mean) and our sampling algorithm. Beyond the motivating example and in order to allow for the use of a diverse range of stochastic optimizers, we present our results using an $L^2$-accurate score estimation assumption, which crucially is formed under an expectation with respect to the stochastic optimizer and our novel auxiliary process that uses only known information. This approach yields the best known convergence rate for our sampling algorithm.

In this paper, we introduce DiffusionMat, a novel image matting framework that employs a diffusion model for the transition from coarse to refined alpha mattes. Diverging from conventional methods that utilize trimaps merely as loose guidance for alpha matte prediction, our approach treats image matting as a sequential refinement learning process. This process begins with the addition of noise to trimaps and iteratively denoises them using a pre-trained diffusion model, which incrementally guides the prediction towards a clean alpha matte. The key innovation of our framework is a correction module that adjusts the output at each denoising step, ensuring that the final result is consistent with the input image’s structures. We also introduce the Alpha Reliability Propagation, a novel technique designed to maximize the utility of available guidance by selectively enhancing the trimap regions with confident alpha information, thus simplifying the correction task. To train the correction module, we devise specialized loss functions that target the accuracy of the alpha matte’s edges and the consistency of its opaque and transparent regions. We evaluate our model across several image matting benchmarks, and the results indicate that DiffusionMat consistently outperforms existing methods. Project page at~\url{https://cnnlstm.github.io/DiffusionMat

Diffusion generative models have emerged as a powerful framework for addressing problems in structural biology and structure-based drug design. These models operate directly on 3D molecular structures. Due to the unfavorable scaling of graph neural networks (GNNs) with graph size as well as the relatively slow inference speeds inherent to diffusion models, many existing molecular diffusion models rely on coarse-grained representations of protein structure to make training and inference feasible. However, such coarse-grained representations discard essential information for modeling molecular interactions and impair the quality of generated structures. In this work, we present a novel GNN-based architecture for learning latent representations of molecular structure. When trained end-to-end with a diffusion model for de novo ligand design, our model achieves comparable performance to one with an all-atom protein representation while exhibiting a 3-fold reduction in inference time.