Fast constrained sampling in pre-trained diffusion models

Diffusion models have dominated the field of large, generative image models, with the prime examples of Stable Diffusion and DALL-E 3 being widely adopted. These models have been trained to perform text-conditioned generation on vast numbers of image-caption pairs and as a byproduct, have acquired general knowledge about natural image statistics. However, when confronted with the task of constrained sampling, e.g. generating the right half of an image conditioned on the known left half, applying these models is a delicate and slow process, with previously proposed algorithms relying on expensive iterative operations that are usually orders of magnitude slower than text-based inference. This is counter-intuitive, as image-conditioned generation should rely less on the difficult-to-learn semantic knowledge that links captions and imagery, and should instead be achievable by lower-level correlations among image pixels. In practice, inverse models are trained or tuned separately for each inverse problem, e.g. by providing parts of images during training as an additional condition, to allow their application in realistic settings. However, we argue that this is not necessary and propose an algorithm for fast-constrained sampling in large pre-trained diffusion models (Stable Diffusion) that requires no expensive backpropagation operations through the model and produces results comparable even to the state-of-the-art \emph{tuned} models. Our method is based on a novel optimization perspective to sampling under constraints and employs a numerical approximation to the expensive gradients, previously computed using backpropagation, incurring significant speed-ups.

TopoDiffusionNet: A Topology-aware Diffusion Model

Diffusion models excel at creating visually impressive images but often struggle to generate images with a specified topology. The Betti number, which represents the number of structures in an image, is a fundamental measure in topology. Yet, diffusion models fail to satisfy even this basic constraint. This limitation restricts their utility in applications requiring exact control, like robotics and environmental modeling. To address this, we propose TopoDiffusionNet (TDN), a novel approach that enforces diffusion models to maintain the desired topology. We leverage tools from topological data analysis, particularly persistent homology, to extract the topological structures within an image. We then design a topology-based objective function to guide the denoising process, preserving intended structures while suppressing noisy ones. Our experiments across four datasets demonstrate significant improvements in topological accuracy. TDN is the first to integrate topology with diffusion models, opening new avenues of research in this area.

Learning Frame-Wise Emotion Intensity for Audio-Driven Talking-Head Generation

Human emotional expression is inherently dynamic, complex, and fluid, characterized by smooth transitions in intensity throughout verbal communication. However, the modeling of such intensity fluctuations has been largely overlooked by previous audio-driven talking-head generation methods, which often results in static emotional outputs. In this paper, we explore how emotion intensity fluctuates during speech, proposing a method for capturing and generating these subtle shifts for talking-head generation. Specifically, we develop a talking-head framework that is capable of generating a variety of emotions with precise control over intensity levels. This is achieved by learning a continuous emotion latent space, where emotion types are encoded within latent orientations and emotion intensity is reflected in latent norms. In addition, to capture the dynamic intensity fluctuations, we adopt an audio-to-intensity predictor by considering the speaking tone that reflects the intensity. The training signals for this predictor are obtained through our emotion-agnostic intensity pseudo-labeling method without the need of frame-wise intensity labeling. Extensive experiments and analyses validate the effectiveness of our proposed method in accurately capturing and reproducing emotion intensity fluctuations in talking-head generation, thereby significantly enhancing the expressiveness and realism of the generated outputs.

TalkinNeRF: Animatable Neural Fields for Full-Body Talking Humans

We introduce a novel framework that learns a dynamic neural radiance field (NeRF) for full-body talking humans from monocular videos. Prior work represents only the body pose or the face. However, humans communicate with their full body, combining body pose, hand gestures, as well as facial expressions. In this work, we propose TalkinNeRF, a unified NeRF-based network that represents the holistic 4D human motion. Given a monocular video of a subject, we learn corresponding modules for the body, face, and hands, that are combined together to generate the final result. To capture complex finger articulation, we learn an additional deformation field for the hands. Our multi-identity representation enables simultaneous training for multiple subjects, as well as robust animation under completely unseen poses. It can also generalize to novel identities, given only a short video as input. We demonstrate state-of-the-art performance for animating full-body talking humans, with fine-grained hand articulation and facial expressions.

JEAN: Joint Expression and Audio-guided NeRF-based Talking Face Generation

We introduce a novel method for joint expression and audio-guided talking face generation. Recent approaches either struggle to preserve the speaker identity or fail to produce faithful facial expressions. To address these challenges, we propose a NeRF-based network. Since we train our network on monocular videos without any ground truth, it is essential to learn disentangled representations for audio and expression. We first learn audio features in a self-supervised manner, given utterances from multiple subjects. By incorporating a contrastive learning technique, we ensure that the learned audio features are aligned to the lip motion and disentangled from the muscle motion of the rest of the face. We then devise a transformer-based architecture that learns expression features, capturing long-range facial expressions and disentangling them from the speech-specific mouth movements. Through quantitative and qualitative evaluation, we demonstrate that our method can synthesize high-fidelity talking face videos, achieving state-of-the-art facial expression transfer along with lip synchronization to unseen audio.

Shadow Removal Refinement via Material-Consistent Shadow Edges

Shadow boundaries can be confused with material boundaries as both exhibit sharp changes in luminance or contrast within a scene. However, shadows do not modify the intrinsic color or texture of surfaces. Therefore, on both sides of shadow edges traversing regions with the same material, the original color and textures should be the same if the shadow is removed properly. These shadow/shadow-free pairs are very useful but hard-to-collect supervision signals. The crucial contribution of this paper is to learn how to identify those shadow edges that traverse material-consistent regions and how to use them as self-supervision for shadow removal refinement during test time. To achieve this, we fine-tune SAM, an image segmentation foundation model, to produce a shadow-invariant segmentation and then extract material-consistent shadow edges by comparing the SAM segmentation with the shadow mask. Utilizing these shadow edges, we introduce color and texture-consistency losses to enhance the shadow removal process. We demonstrate the effectiveness of our method in improving shadow removal results on more challenging, in-the-wild images, outperforming the state-of-the-art shadow removal methods. Additionally, we propose a new metric and an annotated dataset for evaluating the performance of shadow removal methods without the need for paired shadow/shadow-free data.

Look Hear: Gaze Prediction for Speech-directed Human Attention

For computer systems to effectively interact with humans using spoken language, they need to understand how the words being generated affect the users' moment-by-moment attention. Our study focuses on the incremental prediction of attention as a person is seeing an image and hearing a referring expression defining the object in the scene that should be fixated by gaze. To predict the gaze scanpaths in this incremental object referral task, we developed the Attention in Referral Transformer model or ART, which predicts the human fixations spurred by each word in a referring expression. ART uses a multimodal transformer encoder to jointly learn gaze behavior and its underlying grounding tasks, and an autoregressive transformer decoder to predict, for each word, a variable number of fixations based on fixation history. To train ART, we created RefCOCO-Gaze, a large-scale dataset of 19,738 human gaze scanpaths, corresponding to 2,094 unique image-expression pairs, from 220 participants performing our referral task. In our quantitative and qualitative analyses, ART not only outperforms existing methods in scanpath prediction, but also appears to capture several human attention patterns, such as waiting, scanning, and verification.

Assessing Sample Quality via the Latent Space of Generative Models

Advances in generative models increase the need for sample quality assessment. To do so, previous methods rely on a pre-trained feature extractor to embed the generated samples and real samples into a common space for comparison. However, different feature extractors might lead to inconsistent assessment outcomes. Moreover, these methods are not applicable for domains where a robust, universal feature extractor does not yet exist, such as medical images or 3D assets. In this paper, we propose to directly examine the latent space of the trained generative model to infer generated sample quality. This is feasible because the quality a generated sample directly relates to the amount of training data resembling it, and we can infer this information by examining the density of the latent space. Accordingly, we use a latent density score function to quantify sample quality. We show that the proposed score correlates highly with the sample quality for various generative models including VAEs, GANs and Latent Diffusion Models. Compared with previous quality assessment methods, our method has the following advantages: 1) pre-generation quality estimation with reduced computational cost, 2) generalizability to various domains and modalities, and 3) applicability to latent-based image editing and generation methods. Extensive experiments demonstrate that our proposed methods can benefit downstream tasks such as few-shot image classification and latent face image editing. Code is available at https://github.com/cvlab-stonybrook/LS-sample-quality.

$\infty$-Brush: Controllable Large Image Synthesis with Diffusion Models in Infinite Dimensions

Synthesizing high-resolution images from intricate, domain-specific information remains a significant challenge in generative modeling, particularly for applications in large-image domains such as digital histopathology and remote sensing. Existing methods face critical limitations: conditional diffusion models in pixel or latent space cannot exceed the resolution on which they were trained without losing fidelity, and computational demands increase significantly for larger image sizes. Patch-based methods offer computational efficiency but fail to capture long-range spatial relationships due to their overreliance on local information. In this paper, we introduce a novel conditional diffusion model in infinite dimensions, $\infty$-Brush for controllable large image synthesis. We propose a cross-attention neural operator to enable conditioning in function space. Our model overcomes the constraints of traditional finite-dimensional diffusion models and patch-based methods, offering scalability and superior capability in preserving global image structures while maintaining fine details. To our best knowledge, $\infty$-Brush is the first conditional diffusion model in function space, that can controllably synthesize images at arbitrary resolutions of up to $4096\times4096$ pixels. The code is available at https://github.com/cvlab-stonybrook/infinity-brush.

Weighting Pseudo-Labels via High-Activation Feature Index Similarity and Object Detection for Semi-Supervised Segmentation

Semi-supervised semantic segmentation methods leverage unlabeled data by pseudo-labeling them. Thus the success of these methods hinges on the reliablility of the pseudo-labels. Existing methods mostly choose high-confidence pixels in an effort to avoid erroneous pseudo-labels. However, high confidence does not guarantee correct pseudo-labels especially in the initial training iterations. In this paper, we propose a novel approach to reliably learn from pseudo-labels. First, we unify the predictions from a trained object detector and a semantic segmentation model to identify reliable pseudo-label pixels. Second, we assign different learning weights to pseudo-labeled pixels to avoid noisy training signals. To determine these weights, we first use the reliable pseudo-label pixels identified from the first step and labeled pixels to construct a prototype for each class. Then, the per-pixel weight is the structural similarity between the pixel and the prototype measured via rank-statistics similarity. This metric is robust to noise, making it better suited for comparing features from unlabeled images, particularly in the initial training phases where wrong pseudo labels are prone to occur. We show that our method can be easily integrated into four semi-supervised semantic segmentation frameworks, and improves them in both Cityscapes and Pascal VOC datasets.