CVPR 2025 Papers

Diffusion models with transformer architectures have demonstrated promising capabilities in generating high-fidelity images and scalability for high resolution. However, iterative sampling process required for synthesis is very resource-intensive. A line of work has focused on distilling solutions to probability flow ODEs into few-step student models. Nevertheless, existing methods have been limited by their reliance on the most recent denoised samples as input, rendering them susceptible to exposure bias. To address this limitation, we propose AutoRegressive Distillation (ARD), a novel approach that leverages the historical trajectory of the ODE to predict future steps. ARD offers two key benefits: 1) it mitigates exposure bias by utilizing a predicted historical trajectory that is less susceptible to accumulated errors, and 2) it leverages the previous history of the ODE trajectory as a more effective source of coarse-grained information. ARD modifies the teacher transformer architecture by adding token-wise time embedding to mark each input from the trajectory history and employs a block-wise causal attention mask for training. Furthermore, incorporating historical inputs only in lower transformer layers enhances performance and efficiency. We validate the effectiveness of ARD in a class-conditioned generation on ImageNet and T2I synthesis. Our model achieves a 5xreduction in FID degradation compared to the baseline methods while requiring only 1.1% extra FLOPs on ImageNet-256. Moreover, ARD reaches FID of 1.84 on ImageNet-256 in merely 4 steps and outperforms the publicly available 1024p text-to-image distilled models in prompt adherence score with a minimal drop in FID compared to the teacher. Our code is available at https://github.com/alsdudrla10/ARD.

Designing structured visuals such as presentation slides is essential for communicative needs, necessitating both content creation and visual planning skills. In this work, we tackle the challenge of automated slide generation, where models produce slide presentations from natural language (NL) instructions. We first introduce the SlidesBench benchmark, the first benchmark for slide generation with 7k training and 585 testing examples derived from 310 slide decks across 10 domains. SlidesBench supports evaluations that are (i) reference-based to measure similarity to a target slide, and (ii) reference-free to measure the design quality of generated slides alone. We benchmark end-to-end image generation and program generation methods with a variety of models, and find that programmatic methods produce higher-quality slides in user-interactable formats. Built on the success of program generation, we create AutoPresent, an 8B Llama-based model trained on 7k pairs of instructions paired with code for slide generation, and achieve results comparable to the closed-source model GPT-4o. We further explore iterative design refinement where the model is tasked to self-refine its own output, and we found that this process improves the slide's quality. We hope that our work will provide a basis for future work on generating structured visuals.

This work tackles the fidelity objective in the perceptual super-resolution (SR) task. Specifically, we address the shortcomings of pixel-level \mathcal L _\text p loss (\mathcal L _\text pix ) in the GAN-based SR framework. Since \mathcal L _\text pix is known to have a trade-off relationship against perceptual quality, prior methods often multiply a small scale factor or utilize low-pass filters. However, this work shows that these circumventions fail to address the fundamental factor that induces blurring. Accordingly, we focus on two points: 1) precisely discriminating the subcomponent of \mathcal L _\text pix that contributes to blurring, and 2) guiding reconstruction only based on the factor that is free from this trade-off relationship. We show that this can be achieved in a surprisingly simple manner, with an Auto-Encoder (AE) pretrained using \mathcal L _\text pix . Based on this insight, we propose the Auto-Encoded Supervision for Optimal Penalization loss (\mathcal L _\text AESOP ), a novel loss function that measures distance in the AE space (the space after the decoder, not the bottleneck), rather than in the raw pixel space. By simply substituting \mathcal L _\text pix with \mathcal L _\text AESOP , we can provide effective reconstruction guidance without compromising perceptual quality. Designed for simplicity, our method enables easy integration into existing SR frameworks. Extensive experiments demonstrate the effectiveness of AESOP.

Super-resolution (SR), a classical inverse problem in computer vision, is inherently ill-posed, inducing a distribution of plausible solutions for every input. However, the desired result is not simply the expectation of this distribution, which is the blurry image obtained by minimizing pixelwise error, but rather the sample with the highest image quality. A variety of techniques, from perceptual metrics to adversarial losses, are employed to this end. In this work, we explore an alternative: utilizing powerful non-reference image quality assessment (NR-IQA) models in the SR context. We begin with a comprehensive analysis of NR-IQA metrics on human-derived SR data, identifying both the accuracy (human alignment) and complementarity of different metrics. Then, we explore two methods of applying NR-IQA models to SR learning: (i) altering data sampling, by building on an existing multi-ground-truth SR framework, and (ii) directly optimizing a differentiable quality score. Our results demonstrate a more human-centric perception-distortion tradeoff, focusing less on non-perceptual pixelwise distortion, instead improving the balance between perceptual fidelity and human-tuned NR-IQA measures.

Designing 3D scenes is traditionally a challenging and laborious task that demands both artistic expertise and proficiency with complex software. Recent advances in text-to-3D generation have greatly simplified this process by letting users create scenes based on simple text descriptions. However, as these methods generally require extra training or in-context learning, their performance is often hindered by the limited availability of high-quality 3D data. In contrast, modern text-to-image models learned from web-scale images can generate scenes with diverse, reliable spatial layouts and consistent, visually appealing styles. Our key insight is that instead of learning directly from 3D scenes, we can leverage generated 2D images as an intermediary to guide 3D synthesis. In light of this, we introduce ArtiScene, a training-free automated pipeline for scene design that integrates the flexibility of free-form text-to-image generation with the diversity and reliability of 2D intermediary layouts. We generate the 2D intermediary image from a scene description, extract object shapes and appearances, create 3D models, and assemble them into the final scene with geometry, position, and pose extracted from the same image. Being generalizable to a wide range of scenes and styles, ArtiScene is shown to outperform state-of-the-art benchmarks by a large margin in layout and aesthetic quality by quantitative metrics. It also averages a 74.89% winning rate in extensive user studies and 95.07% in GPT evaluation.

Subject-driven text-to-image generation has demonstrated remarkable advancements in its ability to learn and capture characteristics of a subject using only a limited number of images. However, existing methods commonly rely on high-quality images for training and often struggle to generate reasonable images when the input images are blemished by artifacts. This is primarily attributed to the inadequate capability of current techniques in distinguishing subject-related features from disruptive artifacts. In this paper, we introduce ArtiFade to tackle this issue and successfully generate high-quality artifact-free images from blemished datasets. Specifically, ArtiFade exploits fine-tuning of a pre-trained text-to-image model, aiming to remove artifacts. The elimination of artifacts is achieved by utilizing a specialized dataset that encompasses both unblemished images and their corresponding blemished counterparts during fine-tuning. ArtiFade also ensures the preservation of the original generative capabilities inherent within the diffusion model, thereby enhancing the overall performance of subject-driven methods in generating high-quality and artifact-free images. We further devise evaluation benchmarks tailored for this task. Through extensive qualitative and quantitative experiments, we demonstrate the generalizability of ArtiFade in effective artifact removal under both in-distribution and out-of-distribution scenarios.

This paper presents a novel framework for modeling and conditional generation of 3D articulated objects. Troubled by flexibility-quality tradeoffs, existing methods are often limited to using predefined structures or retrieving shapes from static datasets. To address these challenges, we parameterize an articulated object as a tree of tokens and employ a transformer to generate both the object's high-level geometry code and its kinematic relations. Subsequently, each sub-part's geometry is further decoded using a signed-distance-function (SDF) shape prior, facilitating the synthesis of high-quality 3D shapes. Our approach enables the generation of diverse objects with high-quality geometry and varying number of parts. Comprehensive experiments on conditional generation from text descriptions demonstrate the effectiveness and flexibility of our method.

The ultimate goal of generative models is to perfectly capture the data distribution. For image generation, common metrics of visual quality (e.g., FID) and the perceived truthfulness of generated images seem to suggest that we are nearing this goal. However, through distribution classification tasks, we reveal that, from the perspective of neural network-based classifiers, even advanced diffusion models are still far from this goal. Specifically, classifiers are able to consistently and effortlessly distinguish real images from generated ones across various settings. Moreover, we uncover an intriguing discrepancy: classifiers can easily differentiate between diffusion models with comparable performance (e.g., U-ViT-H vs. DiT-XL), but struggle to distinguish between models within the same family but of different scales (e.g., EDM2-XS vs. EDM2-XXL). Our methodology carries several important implications. First, it naturally serves as a diagnostic tool for diffusion models by analyzing specific features of generated data. Second, it sheds light on the model autophagy disorder and offers insights into the use of generated data: augmenting real data with generated data is more effective than replacing it. Third, classifier guidance can significantly enhance the realism of generated images.

We introduce ArcPro, a novel learning framework built on architectural programs to recover structured 3D abstractions from highly sparse and low-quality point clouds. Specifically, we design a domain-specific language (DSL) to hierarchically represent building structures as a program, which can be efficiently converted into a mesh. We bridge feedforward and inverse procedural modeling by using a feedforward process for training data synthesis, allowing the network to make reverse predictions. We train an encoder-decoder on the points-program pairs to establish a mapping from unstructured point clouds to architectural programs, where a 3D convolutional encoder extracts point cloud features and a transformer decoder autoregressively predicts the programs in a tokenized form. Inference by our method is highly efficient and produces plausible and faithful 3D abstractions. Comprehensive experiments demonstrate that ArcPro outperforms both traditional architectural proxy reconstruction and learning-based abstraction methods. We further explore its potential when working with multi-view image and natural language inputs.

Inspired by the effectiveness of 3D Gaussian Splatting (3DGS) in reconstructing detailed 3D scenes within multi-view setups and the emergence of large 2D human foundation models, we introduce Arc2Avatar, the first SDS-based method utilizing a human face foundation model as guidance with just a single image as input. To achieve that, we extend such a model for diverse-view human head generation by fine-tuning on synthetic data and modifying its conditioning. Our avatars maintain a dense correspondence with a human face mesh template, allowing blendshape-based expression generation. This is achieved through a modified 3DGS approach, connectivity regularizers, and a strategic initialization tailored for our task. Additionally, we propose an optional efficient SDS-based correction step to refine the blendshape expressions, enhancing realism and diversity. Experiments demonstrate that Arc2Avatar achieves state-of-the-art realism and identity preservation, effectively addressing color issues by allowing the use of very low guidance, enabled by our strong identity prior and initialization strategy, without compromising detail.

This study presents a new image super-resolution (SR) technique based on diffusion inversion, aiming at harnessing the rich image priors encapsulated in large pre-trained diffusion models to improve SR performance. We design a Partial noise Prediction strategy to construct an intermediate state of the diffusion model, which serves as the starting sampling point. Central to our approach is a deep noise predictor to estimate the optimal noise maps for the forward diffusion process. Once trained, this noise predictor can be used to initialize the sampling process partially along the diffusion trajectory, generating the desirable high-resolution result. Compared to existing approaches, our method offers a flexible and efficient sampling mechanism that supports an arbitrary number of sampling steps, ranging from one to five. Even with a single sampling step, our method demonstrates superior or comparable performance to recent state-of-the-art approaches. The code and model will be made publicly available.

Recent text-to-3D generation models have demonstrated remarkable abilities in producing high-quality 3D assets. Despite their great advancements, current models struggle to generate satisfying 3D objects with complex attributes. The difficulty for such complex attributes 3D generation arises from two aspects: (1) existing text-to-3D approaches typically lift text-to-image models to extract semantics via text encoders, while the text encoder exhibits limited comprehension ability for long descriptions, leading to deviated cross-attention focus, subsequently wrong attribute binding in generated results. (2) Objects with complex attributes often exhibit occlusion relationships between different parts, which demands a reasonable generation order as well as explicit disentanglement of different parts to enable structural coherent and attribute following results. Though some works introduce manual efforts to alleviate the above issues, their quality is unstable and highly reliant on manual information. To tackle above problems, we propose a automated method Hierarchical-Chain-of-Generation (HCoG). It leverages a large language model to analyze the long description, decomposes it into several blocks representing different object parts, and organizes an optimal generation order from in to out according to the occlusion relationship between parts, turning the whole generation process into a hierarchical chain. For optimization within each block, we first generate the necessary components coarsely, then bind their attributes precisely by target region localization and corresponding 3D Gaussian kernel optimization. For optimization between blocks, we introduce Gaussian Extension and Label Elimination to seamlessly generate new parts by extending new Gaussian kernels, re-assigning semantic labels, and eliminating unnecessary kernels, ensuring that only relevant parts are added without disrupting previously optimized parts. Experiments validate HCoG's effectiveness in handling complex attributes 3D assets and witnesses high-quality results. The code is available at https://github.com/Wakals/GASCOL.

Instruction-based image editing aims to modify specific image elements with natural language instructions. However, current models in this domain often struggle to execute complex user instructions accurately, as they are trained on low-quality data with limited editing types. We present AnyEdit, a comprehensive multi-modal instruction editing dataset, comprising 2.5 million high-quality editing pairs spanning over 20 editing types and five domains. We ensure the diversity and quality of the AnyEdit collection through three aspects: initial data diversity, adaptive editing process, and automated selection of editing results. Using the dataset, we further train a novel AnyEdit Stable Diffusion with task-aware routing and learnable task embedding for unified image editing. Comprehensive experiments on three benchmark datasets show that AnyEdit consistently boosts the performance of diffusion-based editing models. This presents prospects for developing instruction-driven image editing models that support human creativity.

Recent works have established that AI models introduce spectral artifacts into generated images and propose approaches for learning to capture them using labeled data. However, the significant differences in such artifacts among different generative models hinder these approaches from generalizing to generators not seen during training. In this work, we build upon the key idea that the spectral distribution of real images constitutes both an invariant and highly discriminative pattern for AI-generated image detection. To model this under a self-supervised setup, we employ masked spectral learning using the pretext task of frequency reconstruction. Since generated images constitute out-of-distribution samples for this model, we propose spectral reconstruction similarity to capture this divergence. Moreover, we introduce spectral context attention, which enables our approach to efficiently capture subtle spectral inconsistencies in images of any resolution. Our spectral AI-generated image detection approach (SPAI) achieves a 5.5% absolute improvement in AUC over the previous state-of-the-art across 13 recent generative approaches, while exhibiting robustness against common online perturbations. Code is available on https://mever-team.github.io/spai.

Despite the remarkable progress of deep learning-based methods in medical image segmentation, their use in clinical practice remains limited for two main reasons. First, obtaining a large medical dataset with precise annotations to train segmentation models is challenging. Secondly, most current segmentation techniques generate a single deterministic segmentation mask for each image. However, in real-world scenarios, there is often significant uncertainty regarding what defines the "correct" segmentation, and various expert annotators might provide different segmentations for the same image. To tackle both of these problems, we propose Annotation Ambiguity Aware Semi-Supervised Medical Image Segmentation (AmbiSSL). AmbiSSL combines a small amount of multi-annotator labeled data and a large set of unlabeled data to generate diverse and plausible segmentation maps. Our method consists of three key components: (1) The Diverse Pseudo-Label Generation (DPG) module utilizes multiple decoders, created by performing randomized pruning on the original backbone decoder. These pruned decoders enable the generation of a diverse pseudo-label set; (2) a Semi-Supervised Latent Distribution Learning (SSLDL) module constructs a common latent space by utilizing both ground truth annotations and pseudo-label set; and (3) a Cross-Decoder Supervision (CDS) module, which enables pruned decoders to guide each other's learning. We evaluated the proposed method on two publicly available datasets. Extensive experiments demonstrate that AmbiSSL can generate diverse segmentation maps using only a small amount of labeled data and abundant unlabeled data, offering a more practical solution for medical image segmentation by reducing reliance on large labeled datasets.

We propose a unified approach for video-controlled generation, enabling text-based guidance and manual annotations to control the generation of videos, similar to camera direction guidance. Specifically, we designed a two-stage algorithm. In the first stage, we convert all control information into frame-by-frame motion flows. In the second stage, we use these motion flows as guidance to control the final video generation. Additionally, to reduce instability in the generated videos caused by large motion variations (such as those from camera movement, object motion, or manual inputs), which can result in flickering or the intermittent disappearance of objects, we transform the temporal feature computation in the video model into frequency-domain feature computation. This is because frequency-domain signals better capture the essential characteristics of an image, and by ensuring consistency in the video's frequency-domain features, we can enhance temporal coherence and reduce flickering in the final generated video.

This paper addresses a promising yet underexplored task, Image-to-Sounding-Video (I2SV) generation, which animates a static image and generates synchronized sound simultaneously. Despite advances in video and audio generation models, challenges remain to develop a unified model for generating naturally sounding videos. In this work, we propose a novel approach that leverages two separate pretrained diffusion models and makes vision and audio influence each other during generation based on the Diffusion Transformer (DiT) architecture. First, the individual video and audio pretrained generation models are decomposed into input, output, and expert sub-modules. We propose using a unified joint DiT block to integrate the expert sub-modules to effectively model the interaction between the two modalities, resulting in high-quality I2SV generation. Then, we introduce a joint classifier-free guidance technique to boost the performance during joint generation. Finally, we conduct extensive experiments on three popular benchmark datasets, and in both objective and subjective evaluation our method surpass all the baseline methods in almost all metrics. Case studies show our generated sounding videos are high quality and synchronized between video and audio.

Text-driven motion generation has made significant strides in recent years. However, most existing works focus on human motion, largely overlooking the rich and diverse behaviors of animals. Understanding and synthesizing animal motion have important applications in wildlife conservation, animal ecology, and biomechanics. Animal motion modeling presents unique challenges due to species diversity, varied morphological structures, and different behavioral patterns in response to similar textual descriptions. To address these challenges, we propose AniMo for text-driven animal motion generation. AniMo consists of two stages: motion tokenization and text-to-motion generation. In the motion tokenization stage, we encode motions using a joint-aware spatiotemporal encoder with species-aware feature modulation, enabling the model to adapt to diverse skeletal structures across species. In the text-to-motion generation stage, we employ masked modeling to jointly learn the mappings between textual descriptions and motion tokens. Additionally, we introduce AniMo4D, a large-scale dataset containing 78,149 motion sequences and 185,435 textual descriptions across 114 animal species. Experimental results show that AniMo achieves superior performance on both the AniMo4D and AnimalML3D datasets, effectively capturing diverse morphological structures and behavioral patterns across animal species.