ICLR 2024 论文列表

Image-to-image (I2I) translation is vital in computer vision tasks like style transfer and domain adaptation. While recent advances in GAN have enabled high-quality sample generation, real-world challenges such as noise and distortion remain significant obstacles. Although Gaussian noise injection during training has been utilized, its theoretical underpinnings have been unclear. This work provides a robust theoretical framework elucidating the role of Gaussian noise injection in I2I translation models. We address critical questions on the influence of noise variance on distribution divergence, resilience to unseen noise types, and optimal noise intensity selection. Our contributions include connecting $f$-divergence and score matching, unveiling insights into the impact of Gaussian noise on aligning probability distributions, and demonstrating generalized robustness implications. We also explore choosing an optimal training noise level for consistent performance in noisy environments. Extensive experiments validate our theoretical findings, showing substantial improvements over various I2I baseline models in noisy settings. Our research rigorously grounds Gaussian noise injection for I2I translation, offering a sophisticated theoretical understanding beyond heuristic applications.

The current advancements in open domain text generation have been spearheaded by Transformer-based large language models. Leveraging efficient parallelization and vast training datasets, these models achieve unparalleled text generation capabilities. Even so, current models are known to suffer from deficiencies such as repetitive texts, looping issues, and lack of robustness. While adversarial training through generative adversarial networks (GAN) is a proposed solution, earlier research in this direction has predominantly focused on older architectures, or narrow tasks. As a result, this approach is not yet compatible with modern language models for open-ended text generation, leading to diminished interest within the broader research community. We propose a computationally efficient GAN approach for sequential data that utilizes the parallelization capabilities of Transformer models. Our method revolves around generating multiple branching sequences from each training sample, while also incorporating the typical next-step prediction loss on the original data. In this way, we achieve a dense reward and loss signal for both the generator and the discriminator, resulting in a stable training dynamic. We apply our training method to pre-trained language models, using data from their original training set but less than 0.01% of the available data. A comprehensive human evaluation shows that our method significantly improves the quality of texts generated by the model while avoiding the previously reported sparsity problems of GAN approaches. Even our smaller models outperform larger original baseline models with more than 16 times the number of parameters. Finally, we corroborate previous claims that perplexity on held-out data is not a sufficient metric for measuring the quality of generated texts.

Recent advancements in diffusion models have significantly enhanced the data synthesis with 2D control. Yet, precise 3D control in street view generation, crucial for 3D perception tasks, remains elusive. Specifically, utilizing Bird's-Eye View (BEV) as the primary condition often leads to challenges in geometry control (e.g., height), affecting the representation of object shapes, occlusion patterns, and road surface elevations, all of which are essential to perception data synthesis, especially for 3D object detection tasks. In this paper, we introduce MagicDrive, a novel street view generation framework, offering diverse 3D geometry controls including camera poses, road maps, and 3D bounding boxes, together with textual descriptions, achieved through tailored encoding strategies. Besides, our design incorporates a cross-view attention module, ensuring consistency across multiple camera views. With MagicDrive, we achieve high-fidelity street-view image & video synthesis that captures nuanced 3D geometry and various scene descriptions, enhancing tasks like BEV segmentation and 3D object detection. Project Website: https://flymin.github.io/magicdrive

Generating a set of text is a common challenge for many NLP applications, for example, automatically providing multiple keyphrases for a document to facilitate user reading. Existing generative models use a sequential decoder that generates a single sequence successively, and the set generation problem is converted to sequence generation via concatenating multiple text into a long text sequence. However, the elements of a set are unordered, which makes this scheme suffer from biased or conflicting training signals. In this paper, we propose a branching decoder, which can generate a dynamic number of tokens at each time-step and branch multiple generation paths. In particular, paths are generated individually so that no order dependence is required. Moreover, multiple paths can be generated in parallel which greatly reduces the inference time. Experiments on several keyphrase generation datasets demonstrate that the branching decoder is more effective and efficient than the existing sequential decoder.

Generative adversarial networks (GANs) have made great success in image inpainting yet still have difficulties tackling large missing regions. In contrast, iterative probabilistic algorithms, such as autoregressive and denoising diffusion models, have to be deployed with massive computing resources for decent effect. To achieve high-quality results with low computational cost, we present a novel pixel spread model (PSM) that iteratively employs decoupled probabilistic modeling, combining the optimization efficiency of GANs with the prediction tractability of probabilistic models. As a result, our model selectively spreads informative pixels throughout the image in a few iterations, largely enhancing the completion quality and efficiency. On multiple benchmarks, we achieve new state-of-the-art performance. Our code and models will be publicly available.

Recent work has showcased the significant potential of diffusion models in pose-guided person image synthesis. However, owing to the inconsistency in pose between the source and target images, synthesizing an image with a distinct pose, relying exclusively on the source image and target pose information, remains a formidable challenge. This paper presents Progressive Conditional Diffusion Models (PCDMs) that incrementally bridge the gap between person images under the target and source poses through three stages. Specifically, in the first stage, we design a simple prior conditional diffusion model that predicts the global features of the target image by mining the global alignment relationship between pose coordinates and image appearance. Then, the second stage establishes a dense correspondence between the source and target images using the global features from the previous stage, and an inpainting conditional diffusion model is proposed to further align and enhance the contextual features, generating a coarse-grained person image. In the third stage, we propose a refining conditional diffusion model to utilize the coarsely generated image from the previous stage as a condition, achieving texture restoration and enhancing fine-detail consistency. The three-stage PCDMs work progressively to generate the final high-quality and high-fidelity synthesized image. Both qualitative and quantitative results demonstrate the consistency and photorealism of our proposed PCDMs under challenging scenarios. The code and model will be available at https://github.com/tencent-ailab/PCDMs.

Diffusion models achieved great success in image synthesis, but still face challenges in high-resolution generation. Through the lens of discrete cosine transformation, we find the main reason is that *the same noise level on a higher resolution results in a higher Signal-to-Noise Ratio in the frequency domain*. In this work, we present Relay Diffusion Model (RDM), which transfers a low-resolution image or noise into an equivalent high-resolution one for diffusion model via blurring diffusion and block noise. Therefore, the diffusion process can continue seamlessly in any new resolution or model without restarting from pure noise or low-resolution conditioning. RDM achieves state-of-the-art FID on CelebA-HQ and sFID on ImageNet 256$\times$256, surpassing previous works such as ADM, LDM and DiT by a large margin. All the codes and checkpoints are open-sourced at \url{https://github.com/THUDM/RelayDiffusion}.

Imagining the future trajectory is the key for robots to make sound planning and successfully reach their goals. Therefore, text-conditioned video prediction (TVP) is an essential task to facilitate general robot policy learning. To tackle this task and empower robots with the ability to foresee the future, we propose a sample and computation-efficient model, named Seer, by inflating the pretrained text-to-image (T2I) stable diffusion models along the temporal axis. We enhance the U-Net and language conditioning model by incorporating computation-efficient spatial-temporal attention. Furthermore, we introduce a novel Frame Sequential Text Decomposer module that dissects a sentence's global instruction into temporally aligned sub-instructions, ensuring precise integration into each frame of generation. Our framework allows us to effectively leverage the extensive prior knowledge embedded in pretrained T2I models across the frames. With the adaptable-designed architecture, Seer makes it possible to generate high-fidelity, coherent, and instruction-aligned video frames by fine-tuning a few layers on a small amount of data. The experimental results on Something Something V2 (SSv2), Bridgedata and EpicKitchens-100 datasets demonstrate our superior video prediction performance with around 480-GPU hours versus CogVideo with over 12,480-GPU hours: achieving the 31\% FVD improvement compared to the current SOTA model on SSv2 and 83.7\% average preference in the human evaluation. Our project is available at https://seervideodiffusion.github.io/

Graphic layout generation, a growing research field, plays a significant role in user engagement and information perception. Existing methods primarily treat layout generation as a numerical optimization task, focusing on quantitative aspects while overlooking the semantic information of layout, such as the relationship between each layout element. In this paper, we propose LayoutNUWA, the first model that treats layout generation as a code generation task to enhance semantic information and harness the hidden layout expertise of large language models~(LLMs). Concretely, we develop a Code Instruct Tuning (CIT) approach comprising three interconnected modules: 1) the Code Initialization (CI) module quantifies the numerical conditions and initializes them as HTML code with strategically placed masks; 2) the Code Completion (CC) module employs the formatting knowledge of LLMs to fill in the masked portions within the HTML code; 3) the Code Rendering (CR) module transforms the completed code into the final layout output, ensuring a highly interpretable and transparent layout generation procedure that directly maps code to a visualized layout. We attain significant state-of-the-art performance (even over 50\% improvements compared to previous works) on multiple datasets, showcasing the strong capabilities of LayoutNUWA.

Creating diverse and high-quality 3D assets with an automatic generative model is highly desirable. Despite extensive efforts on 3D generation, most existing works focus on the generation of a single category or a few categories. In this paper, we introduce a diffusion-based feed-forward framework for synthesizing massive categories of real-world 3D objects \textit{with a single generative model}. Notably, there are three major challenges for this large-vocabulary 3D generation: \textbf{a}) the need for expressive yet efficient 3D representation; \textbf{b}) large diversity in geometry and texture across categories; \textbf{c}) complexity in the appearances of real-world objects. To this end, we propose a novel triplane-based 3D-aware \textbf{Diff}usion model with \textbf{T}rans\textbf{F}ormer, \textbf{DiffTF}, for handling challenges via three aspects. \textbf{1}) Considering efficiency and robustness, we adopt a revised triplane representation and improve the fitting speed and accuracy. \textbf{2}) To handle the drastic variations in geometry and texture, we regard the features of all 3D objects as a combination of generalized 3D knowledge and specialized 3D features. To extract generalized 3D knowledge from diverse categories, we propose a novel 3D-aware transformer with shared cross-plane attention. It learns the cross-plane relations across different planes and aggregates the generalized 3D knowledge with specialized 3D features. \textbf{3}) In addition, we devise the 3D-aware encoder/decoder to enhance the generalized 3D knowledge in the encoded triplanes for handling categories with complex appearances. Extensive experiments on ShapeNet and OmniObject3D (over 200 diverse real-world categories) convincingly demonstrate that a single DiffTF model achieves state-of-the-art large-vocabulary 3D object generation performance with large diversity, rich semantics, and high quality. More results are available at https://difftf.github.io/

The advent of large pre-trained models has brought about a paradigm shift in both visual representation learning and natural language processing. However, clustering unlabeled images, as a fundamental and classic machine learning problem, still lacks an effective solution, particularly for large-scale datasets. In this paper, we propose a novel image clustering pipeline that leverages the powerful feature representation of large pre-trained models such as CLIP and cluster images effectively and efficiently at scale. We first developed a novel algorithm to estimate the number of clusters in a given dataset. We then show that the pre-trained features are significantly more structured by further optimizing the rate reduction objective. The resulting features may significantly improve the clustering accuracy, e.g., from 57\% to 66\% on ImageNet-1k. Furthermore, by leveraging CLIP's multimodality bridge between image and text, we develop a simple yet effective self-labeling algorithm that produces meaningful text labels for the clusters. Through extensive experiments, we show that our pipeline works well on standard datasets such as CIFAR-10, CIFAR-100, and ImageNet-1k. It also extends to datasets without predefined labels, such as LAION-Aesthetics and WikiArts.

Large text-to-video models trained on internet-scale data have demonstrated exceptional capabilities in generating high-fidelity videos from arbitrary textual descriptions. However, similar to proprietary language models, large text-to-video models are often black boxes whose weight parameters are not publicly available, posing a significant challenge to adapting these models to specific domains such as robotics, animation, and personalized stylization. Inspired by how a large language model can be prompted to perform new tasks without access to the model weights, we investigate how to adapt a black-box pretrained text-to-video model to a variety of downstream domains without weight access to the pretrained model. In answering this question, we propose \emph{\methodname}, which leverages the score function of a large pretrained video diffusion model as a probabilistic prior to guide the generation of a task-specific small video model. Our experiments show that, by incorporating broad knowledge and fidelity of the pretrained model probabilistically, a small model with as few as 1.25% parameters of the pretrained model can generate high-quality yet domain-specific videos for a variety of downstream domains such as animation, egocentric modeling, and modeling of simulated and real-world robotics data. As large text-to-video models starting to become available as a service similar to large language models, we advocate for private institutions to expose scores of video diffusion models as outputs in addition to generated videos to allow flexible adaptation of large pretrained text-to-video models by the general public.

Addressing the limitations of text as a source of accurate layout representation in text-conditional diffusion models, many works incorporate additional signals to condition certain attributes within a generated image. Although successful, previous works do not account for the specific localization of said attributes extended into the three dimensional plane. In this context, we present a conditional diffusion model that integrates control over three-dimensional object placement with disentangled representations of global stylistic semantics from multiple exemplar images. Specifically, we first introduce depth disentanglement training to leverage the relative depth of objects as an estimator, allowing the model to identify the absolute positions of unseen objects through the use of synthetic image triplets. We also introduce soft guidance, a method for imposing global semantics onto targeted regions without the use of any additional localization cues. Our integrated framework, Compose and Conquer (CnC), unifies these techniques to localize multiple conditions in a disentangled manner. We demonstrate that our approach allows perception of objects at varying depths while offering a versatile framework for composing localized objects with different global semantics.

Transformer architectures have exhibited remarkable performance in image super-resolution (SR). Since the quadratic computational complexity of the self-attention (SA) in Transformer, existing methods tend to adopt SA in a local region to reduce overheads. However, the local design restricts the global context exploitation, which is crucial for accurate image reconstruction. In this work, we propose the Recursive Generalization Transformer (RGT) for image SR, which can capture global spatial information and is suitable for high-resolution images. Specifically, we propose the recursive-generalization self-attention (RG-SA). It recursively aggregates input features into representative feature maps, and then utilizes cross-attention to extract global information. Meanwhile, the channel dimensions of attention matrices ($query$, $key$, and $value$) are further scaled to mitigate the redundancy in the channel domain. Furthermore, we combine the RG-SA with local self-attention to enhance the exploitation of the global context, and propose the hybrid adaptive integration (HAI) for module integration. The HAI allows the direct and effective fusion between features at different levels (local or global). Extensive experiments demonstrate that our RGT outperforms recent state-of-the-art methods quantitatively and qualitatively. Code and pre-trained models are available at https://github.com/zhengchen1999/RGT.

Data mixing augmentation has been widely applied to improve the generalization ability of deep neural networks. Recently, offline data mixing augmentation, e.g. handcrafted and saliency information-based mixup, has been gradually replaced by automatic mixing approaches. Through minimizing two sub-tasks, namely, mixed sample generation and mixup classification in an end-to-end way, AutoMix significantly improves accuracy on image classification tasks. However, as the optimization objective is consistent for the two sub-tasks, this approach is prone to generating consistent instead of diverse mixed samples, which results in overfitting for target task training. In this paper, we propose AdAutomixup, an adversarial automatic mixup augmentation approach that generates challenging samples to train a robust classifier for image classification, by alternatively optimizing the classifier and the mixup sample generator. AdAutomixup comprises two modules, a mixed example generator, and a target classifier. The mixed sample generator aims to produce hard mixed examples to challenge the target classifier, while the target classifier's aim is to learn robust features from hard mixed examples to improve generalization. To prevent the collapse of the inherent meanings of images, we further introduce an exponential moving average (EMA) teacher and cosine similarity to train AdAutomixup in an end-to-end way. Extensive experiments on seven image benchmarks consistently prove that our approach outperforms the state of the art in various classification scenarios. The source code is available at https://github.com/JinXins/Adversarial-AutoMixup.

Vanilla image completion approaches exhibit sensitivity to large missing regions, attributed to the limited availability of reference information for plausible generation. To mitigate this, existing methods incorporate the extra cue as guidance for image completion. Despite improvements, these approaches are often restricted to employing a *single modality* (e.g., *segmentation* or *sketch* maps), which lacks scalability in leveraging multi-modality for more plausible completion. In this paper, we propose a novel, simple yet effective method for **M**ulti-mod**a**l **G**uided **I**mage **C**ompletion, dubbed **MaGIC**, which not only supports a wide range of single modality as the guidance (e.g., *text*, *canny edge*, *sketch*, *segmentation*, *depth*, and *pose*), but also adapts to arbitrarily customized combinations of these modalities (i.e., *arbitrary multi-modality*) for image completion. For building MaGIC, we first introduce a modality-specific conditional U-Net (MCU-Net) that injects single-modal signal into a U-Net denoiser for single-modal guided image completion. Then, we devise a consistent modality blending (CMB) method to leverage modality signals encoded in multiple learned MCU-Nets through gradient guidance in latent space. Our CMB is *training-free*, thereby avoiding the cumbersome joint re-training of different modalities, which is the secret of MaGIC to achieve exceptional flexibility in accommodating new modalities for completion. Experiments show the superiority of MaGIC over state-of-the-art methods and its generalization to various completion tasks.

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 3.8× speed-ups with the same number of diffusion steps while maintaining high sample quality compared to the baselines.

Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible adversarial perturbations. Furthermore, the concern is exacerbated by the phenomenon that the same adversarial perturbations can fool different task-specific models. Given that VLMs rely on prompts to adapt to different tasks, an intriguing question emerges: Can a single adversarial image mislead all predictions of VLMs when a thousand different prompts are given? This question essentially introduces a novel perspective on adversarial transferability: cross-prompt adversarial transferability. In this work, we propose the Cross-Prompt Attack (CroPA). This proposed method updates the visual adversarial perturbation with learnable textual prompts, which are designed to counteract the misleading effects of the adversarial image. By doing this, CroPA significantly improves the transferability of adversarial examples across prompts. Extensive experiments are conducted to verify the strong cross-prompt adversarial transferability of CroPA with prevalent VLMs including Flamingo, BLIP-2, and InstructBLIP in various different tasks.

Recently, there has been a surge of interest in employing neural networks for graph generation, a fundamental statistical learning problem with critical applications like molecule design and community analysis. However, most approaches encounter significant limitations when generating large-scale graphs. This is due to their requirement to output the full adjacency matrices whose size grows quadratically with the number of nodes. In response to this challenge, we introduce a new, simple, and scalable graph representation named gap encoded edge list (GEEL) that has a small representation size that aligns with the number of edges. In addition, GEEL significantly reduces the vocabulary size by incorporating the gap encoding and bandwidth restriction schemes. GEEL can be autoregressively generated with the incorporation of node positional encoding, and we further extend GEEL to deal with attributed graphs by designing a new grammar. Our findings reveal that the adoption of this compact representation not only enhances scalability but also bolsters performance by simplifying the graph generation process. We conduct a comprehensive evaluation across ten non-attributed and two molecular graph generation tasks, demonstrating the effectiveness of GEEL.