Diffusion models have achieved remarkable success in generating realistic images but suffer from generating accurate human hands, such as incorrect finger counts or irregular shapes. This difficulty arises from the complex task of learning the physical structure and pose of hands from training images, which involves extensive deformations and occlusions. For correct hand generation, our paper introduces a lightweight post-processing solution called HandRefiner. HandRefiner employs a conditional inpainting approach to rectify malformed hands while leaving other parts of the image untouched. We leverage the hand mesh reconstruction model that consistently adheres to the correct number of fingers and hand shape, while also being capable of fitting the desired hand pose in the generated image. Given a generated failed image due to malformed hands, we utilize ControlNet modules to re-inject such correct hand information. Additionally, we uncover a phase transition phenomenon within ControlNet as we vary the control strength. It enables us to take advantage of more readily available synthetic data without suffering from the domain gap between realistic and synthetic hands. Experiments demonstrate that HandRefiner can significantly improve the generation quality quantitatively and qualitatively. The code is available at https://github.com/wenquanlu/HandRefiner.
论文检索
输入标题、作者或关键词，从 1,237 篇学术成果中精准定位
FreeEnhance: Tuning-Free Image Enhancement via Content-Consistent Noising-and-Denoising Process
PDF ↗The emergence of text-to-image generation models has led to the recognition that image enhancement, performed as post-processing, would significantly improve the visual quality of the generated images. Exploring diffusion models to enhance the generated images nevertheless is not trivial and necessitates to delicately enrich plentiful details while preserving the visual appearance of key content in the original image. In this paper, we propose a novel framework, namely FreeEnhance, for content-consistent image enhancement using the off-the-shelf image diffusion models. Technically, FreeEnhance is a two-stage process that firstly adds random noise to the input image and then capitalizes on a pre-trained image diffusion model (i.e., Latent Diffusion Models) to denoise and enhance the image details. In the noising stage, FreeEnhance is devised to add lighter noise to the region with higher frequency to preserve the high-frequent patterns (e.g., edge, corner) in the original image. In the denoising stage, we present three target properties as constraints to regularize the predicted noise, enhancing images with high acutance and high visual quality. Extensive experiments conducted on the HPDv2 dataset demonstrate that our FreeEnhance outperforms the state-of-the-art image enhancement models in terms of quantitative metrics and human preference. More remarkably, FreeEnhance also shows higher human preference compared to the commercial image enhancement solution of Magnific AI.
In this work, we introduce DOPRA, a novel approach designed to mitigate hallucinations in multi-modal large language models (MLLMs). Unlike existing solutions that typically involve costly supplementary training data or the integration of external knowledge sources, DOPRA innovatively addresses hallucinations by decoding specific weighted layer penalties and redistribution, offering an economical and effective solution without additional resources. DOPRA is grounded in unique insights into the intrinsic mechanisms controlling hallucinations within MLLMs, especially the models' tendency to over-rely on a subset of summary tokens in the self-attention matrix, neglecting critical image-related information. This phenomenon is particularly pronounced in certain strata. To counteract this over-reliance, DOPRA employs a strategy of weighted overlay penalties and redistribution in specific layers, such as the 12th layer, during the decoding process. Furthermore, DOPRA includes a retrospective allocation process that re-examines the sequence of generated tokens, allowing the algorithm to reallocate token selection to better align with the actual image content, thereby reducing the incidence of hallucinatory descriptions in auto-generated captions. Overall, DOPRA represents a significant step forward in improving the output quality of MLLMs by systematically reducing hallucinations through targeted adjustments during the decoding process.
Recent text-to-image (T2I) synthesis models have demonstrated intriguing abilities to produce high-quality images based on text prompts. However, current models still face Text-Image Misalignment problem (e.g., attribute errors and relation mistakes) for compositional generation. Existing models attempted to condition T2I models on grounding inputs to improve controllability while ignoring the explicit supervision from the layout conditions. To tackle this issue, we propose Grounded jOint lAyout aLignment (GOAL), an effective framework for T2I synthesis. Two novel modules, discriminative semantic alignment (DSAlign) and masked attention alignment (MAAlign), are proposed and incorporated in this framework to improve the text-image alignment. DSAlign leverages discriminative tasks at the region-wise level to ensure low-level semantic alignment. MAAlign provides high-level attention alignment by guiding the model to focus on the target object. We also build a dataset GOAL2K for model fine-tuning, which composes 2000 semantically accurate image-text pairs and their layout annotations. Comprehensive evaluations on T2I-Compbench, NSR-1K, and Drawbench demonstrate the superior generation performance of our method. Especially, there are improvements of 19%, 13%, and 12% in color, shape, and texture metrics for T2I-Compbench. Additionally, Q-Align metrics demonstrate that our method can generate images of higher quality.
Designing embedding costs is pivotal in modern image steganography. Many studies have shown adjusting symmetric embedding costs to asymmetric ones can enhance steganographic security. However, most existing methods heavily depend on manually defined parameters or rules, limiting security performance improvements. To overcome this limitation, we introduce an advanced GAN-based framework that transitions symmetric costs to asymmetric ones without the need for the manual intervention seen in existing approaches, such as the detailed specification of cost modulation directions and magnitudes. In our framework, we firstly achieve symmetric costs for a cover image, which is randomly split into two sub-images, with part of the secret information embedded into one. Subsequently, we design a GAN model to adjust the embedding costs of the second sub-image to asymmetric, facilitating the secure embedding of the remaining secret information. To support our phased embedding approach, our GAN's discriminator incorporates two steganalyers with different tasks: distinguishing the generator's final output, i.e., the stego image, from both the input cover image and the partially embedded stego image, providing diverse guidance to the generator. In addition, we introduce a simple yet effective update strategy to ensure a stable training process. Comprehensive experiments demonstrate that our method significantly enhances security over existing symmetric steganography techniques, achieving state-of-the-art levels compared to other methods focused on embedding costs adjustments. Additionally, detailed ablation studies validate our approach's effectiveness.
Diffusion models have garnered significant success in generative tasks, emerging as the predominant model in this domain. Despite their success, the substantial computational resources required for training diffusion models restrict their practical applications. In this paper, we resort to the optimal transport theory to accelerate the training of diffusion models, providing an in-depth analysis of the forward diffusion process. It shows that the upper bound on the Wasserstein distance of the distribution between any two timesteps in the diffusion process is an exponential decrease of the initial distance by a factor of times. This finding suggests that the state distribution of the diffusion model has a non-uniform rate of change at different points in time, thus highlighting the different importance of the diffusion timestep. To this end, we propose a novel non-uniform timestep sampling method based on the Bernoulli distribution, which favors more frequent sampling in significant timestep intervals. The key idea is to make the model focus on timesteps with larger differences, thus accelerating the training of the diffusion model. Experiments on benchmark datasets reveal that the proposed method significantly reduces the computational overhead while improving the quality of the generated images.
Speech-driven gesture generation aims at synthesizing a gesture sequence synchronized with the input speech signal. Previous methods leverage neural networks to directly map a compact audio representation to the gesture sequence, ignoring the semantic association of different modalities and failing to deal with salient gestures. In this paper, we propose a novel speech-driven gesture generation method by emphasizing the semantic consistency of salient posture. Specifically, we first learn a joint manifold space for the individual representation of audio and body pose to exploit the inherent semantic association between two modalities, and propose to enforce semantic consistency via a consistency loss. Furthermore, we emphasize the semantic consistency of salient postures by introducing a weakly-supervised detector to identify salient postures, and reweighting the consistency loss to focus more on learning the correspondence between salient postures and the high-level semantics of speech content. In addition, we propose to extract audio features dedicated to facial expression and body gesture separately, and design separate branches for face and body gesture synthesis. Extensive experimental results demonstrate the superiority of our method over the state-of-the-art approaches.
Stylized motion breathes life into characters. However, the fixed skeleton structure and style representation hinder existing data-driven motion synthesis methods from generating stylized motion for various characters. In this work, we propose a generative motion stylization pipeline, named MotionS, for synthesizing diverse and stylized motion on cross-structure characters using cross-modality style prompts. Our key insight is to embed motion style into a cross-modality latent space and perceive the cross-structure skeleton topologies, allowing for motion stylization within a canonical motion space.Specifically, the large-scale Contrastive-Language-Image-Pre-training (CLIP) model is leveraged to construct the cross-modality latent space, enabling flexible style representation within it. Additionally, two topology-encoded tokens are learned to capture the canonical and specific skeleton topologies, facilitating cross-structure topology shifting. Subsequently, the topology-shifted stylization diffusion is designed to generate motion content for the particular skeleton and stylize it in the shifted canonical motion space using multi-modality style descriptions. Through an extensive set of examples, we demonstrate the flexibility and generalizability of our pipeline across various characters and style descriptions. Qualitative and quantitative comparisons show the superiority of our pipeline over state-of-the-arts, consistently delivering high-quality stylized motion across a broad spectrum of skeletal structures.
Recent advancements in Latent Diffusion Models (LDMs) have propelled them to the forefront of various generative tasks. However, their iterative sampling process poses a significant computational burden, resulting in slow generation speeds and limiting their application in text-to-audio generation deployment. In this work, we introduce AudioLCM, a novel consistency-based model tailored for efficient and high-quality text-to-audio generation. Unlike prior approaches that address noise removal through iterative processes, AudioLCM integrates Consistency Models (CMs) into the generation process, facilitating rapid inference through a mapping from any point at any time step to the trajectory's initial point. To overcome the convergence issue inherent in LDMs with reduced sample iterations, we propose the Guided Latent Consistency Distillation with a multi-step Ordinary Differential Equation (ODE) solver. This innovation shortens the time schedule from thousands to dozens of steps while maintaining sample quality, thereby achieving fast convergence and high-quality generation. Furthermore, to optimize the performance of transformer-based neural network architectures, we integrate the advanced techniques pioneered by LLaMA into the foundational framework of transformers. This architecture supports stable and efficient training, ensuring robust performance in text-to-audio synthesis. Experimental results on text-to-audio generation and text-to-music synthesis tasks demonstrate that AudioLCM needs only 2 iterations to synthesize high-fidelity audios, while it maintains sample quality competitive with state-of-the-art models using hundreds of steps. AudioLCM enables a sampling speed of 333x faster than real-time on a single NVIDIA 4090Ti GPU, making generative models practically applicable to text-to-audio generation deployment. Our extensive preliminary analysis shows that each design in AudioLCM is effective. https://AudioLCM.github.io/. Code is Available https://github.com/Text-to-Audio/AudioLCM
Recent progress in large-scale zero-shot speech synthesis has been significantly advanced by language models and diffusion models. However, the generation process of both methods is slow and computationally intensive. Efficient speech synthesis using a lower computing budget to achieve quality on par with previous work remains a significant challenge. In this paper, we present FlashSpeech, a large-scale zero-shot speech synthesis system with approximately 5% of the inference time compared with previous work. FlashSpeech is built on the latent consistency model and applies a novel adversarial consistency training approach that can train from scratch without the need for a pre-trained diffusion model as the teacher. Furthermore, a new prosody generator module enhances the diversity of prosody, making the rhythm of the speech sound more natural. The generation processes of FlashSpeech can be achieved efficiently with one or two sampling steps while maintaining high audio quality and high similarity to the audio prompt for zero-shot speech generation. Our experimental results demonstrate the superior performance of FlashSpeech. Notably, FlashSpeech can be about 20 times faster than other zero-shot speech synthesis systems while maintaining comparable performance in terms of voice quality and similarity. Furthermore, FlashSpeech demonstrates its versatility by efficiently performing tasks like voice conversion, speech editing, and diverse speech sampling. Audio samples can be found in https://flashspeech.github.io/
Is the Text to Motion model robust? Recent advancements in Text to Motion models primarily stem from more accurate predictions of specific actions. However, the text modality typically relies solely on pre-trained Contrastive Language-Image Pretraining (CLIP) models. Our research has uncovered a significant issue with the text-to-motion model: its predictions often exhibit inconsistent outputs, resulting in vastly different or even incorrect poses when presented with semantically similar or identical text inputs. In this paper, we undertake an analysis to elucidate the underlying causes of this instability, establishing a clear link between the unpredictability of model outputs and the erratic attention patterns of the text encoder module. Consequently, we introduce a formal framework aimed at addressing this issue, which we term the Sta ble Text-to-Motion Framework (SATO). SATO consists of three modules, each dedicated to stable attention, stable prediction, and maintaining a balance between accuracy and robustness trade-off. We present a methodology for constructing an SATO that satisfies the stability of attention and prediction. To verify the stability of the model, we introduced a new textual synonym perturbation dataset based on HumanML3D and KIT-ML. Results show that SATO is significantly more stable against synonyms and other slight perturbations while keeping its high accuracy performance. Codes and models are released at https://github.com/sato-team/Stable-Text-to-Motion-Framework.
When ControlNet Meets Inexplicit Masks: A Case Study of ControlNet on its Contour-following Ability
PDF ↗ControlNet excels at creating content that closely matches precise contours in user-provided masks. However, when these masks contain noise, as a frequent occurrence with non-expert users, the output would include unwanted artifacts. This paper first highlights the crucial role of controlling the impact of these inexplicit masks with diverse deterioration levels through in-depth analysis. Subsequently, to enhance controllability with inexplicit masks, an advanced Shape-aware ControlNet consisting of a deterioration estimator and a shape-prior modulation block is devised. The deterioration estimator assesses the deterioration factor of the provided masks. Then this factor is used in a modulation block to adaptively adjust the model's contour-following ability, which helps it dismiss the noise part in the inexplicit masks. Extensive experiments prove its effectiveness in encouraging ControlNet to interpret inaccurate spatial conditions robustly rather than blindly following the given contours, suitable for diverse kinds of conditions. We showcase application scenarios like modifying shape priors and composable shape-controllable generation. Codes are available at github.
Embodied Everyday Task is a popular task in the embodied AI community, requiring agents to make a sequence of actions based on natural language instructions and visual observations. Traditional learning-based approaches face two challenges. Firstly, natural language instructions often lack explicit task planning. Secondly, extensive training is required to equip models with knowledge of the task environment. Previous works based on Large Language Model (LLM) either suffer from poor performance due to the lack of task-specific knowledge or rely on ground truth as few-shot samples. To address the above limitations, we propose a novel approach called Progressive Retrieval Augmented Generation (P-RAG), which not only effectively leverages the powerful language processing capabilities of LLMs but also progressively accumulates task-specific knowledge without ground-truth. Compared to the conventional RAG methods, which retrieve relevant information from the database in a one-shot manner to assist generation, P-RAG introduces an iterative approach to progressively update the database. In each iteration, P-RAG retrieves the latest database and obtains historical information from the previous interaction as experiential references for the current interaction. Moreover, we also introduce a more granular retrieval scheme that not only retrieves similar tasks but also incorporates retrieval of similar situations to provide more valuable reference experiences. Extensive experiments reveal that P-RAG achieves competitive results without utilizing ground truth and can even further improve performance through self-iterations.
Multi-Modal Large Language Models (MM-LLMs) have demonstrated powerful reasoning abilities in various visual question-answering tasks. However, they face the challenge of lacking rigorous reasoning and precise arithmetic, when solving geometry questions. To address this challenge, we propose a novel prompting method, namely Reason-and-Execute (R&E), to enhance the accuracy of solving geometry questions by MM-LLMs. Specifically, the R&E method includes two templates: reasoning template and execution template. We first adopt a reverse-thinking approach to construct a rigorous reasoning template so that it guides MM-LLMs to start reasoning from the most relevant domain knowledge of the question and ultimately identify the arithmetic requirements. We then make use of program-assisted thought to construct execution template in order to guide MM-LLMs to understand the arithmetic requirements from reasoning template and generate executable code block. The answer is finally obtained by executing the code block. We evaluate our prompting method on 9 models in answering questions on 6 datasets (including four geometry datasets and two science datasets) compared to Chain-of-Thought (CoT) and Program-Aided Language (PAL) prompting methods. R&E method shows up to 12.8% improvement compared to CoT and PAL, proving strong reasoning and arithmetic abilities for solving geometry questions of our method. Moreover, we further analyze the answering accuracy from the different perspectives on solving geometric questions, including domain knowledge, geometry shapes, question length, and language. Through multiple analysis, our method is able to enhance the ability of MM-LLMs to solve geometry questions.
We propose Magic Clothing, a latent diffusion model (LDM)-based network architecture for an unexplored garment-driven image synthesis task. Aiming at generating customized characters wearing the target garments with diverse text prompts, the image controllability is the most critical issue, i.e., to preserve the garment details and maintain faithfulness to the text prompts. To this end, we introduce a garment extractor to capture the detailed garment features, and employ self-attention fusion to incorporate them into the pretrained LDMs, ensuring that the garment details remain unchanged on the target character. Then, we leverage the joint classifier-free guidance to balance the control of garment features and text prompts over the generated results. Meanwhile, the proposed garment extractor is a plug-in module applicable to various finetuned LDMs, and it can be combined with other extensions like ControlNet and IP-Adapter to enhance the diversity and controllability of the generated characters. Furthermore, we design Matched-Points-LPIPS (MP-LPIPS), a robust metric for evaluating the consistency of the target image to the source garment. Extensive experiments demonstrate that our Magic Clothing achieves state-of-the-art results under various conditional controls for garment-driven image synthesis. Our source code is available at https://github.com/ShineChen1024/MagicClothing.
Weak diagram analysis abilities of LLMs or Multimodal LLMs greatly limit their application scenarios for scientific academic paper writing. In this work, towards a more versatile copilot for academic paper writing, we mainly focus on strengthening the multi-modal diagram analysis ability of Multimodal LLMs. By parsing Latex source files of academic papers, we carefully build a multi-modal diagram understanding dataset M-Paper. By aligning diagrams in the paper with related paragraphs, we construct professional diagram analysis samples for training and evaluation. M-Paper is the first dataset to support joint comprehension of multiple scientific diagrams, including figures and tables in the format of images or Latex codes. Besides, to better align the copilot with the user's intention, we introduce the 'outline' as the control signal, which could be directly given by the user or revised based on auto-generated ones. Comprehensive experiments with a state-of-the-art Multimodal LLM demonstrate that training on our dataset shows stronger scientific diagram understanding performance. The dataset, code, and model are publicly available at https://github.com/X-PLUG/mPLUG-DocOwl/tree/main/PaperOwl.
Recent advancements in Large Vision-Language Models (VLMs) have underscored their superiority in various multimodal tasks. However, the adversarial robustness of VLMs has not been fully explored. Existing methods mainly assess robustness through unimodal adversarial attacks that perturb images, while assuming inherent resilience against text-based attacks. Different from existing attacks, in this work we propose a more comprehensive strategy that jointly attacks both text and image modalities to exploit a broader spectrum of vulnerability within VLMs. Specifically, we propose a dual optimization objective aimed at guiding the model to generate highly toxic affirmative responses. Our attack method begins by optimizing an adversarial image prefix from random noise to generate diverse harmful responses in the absence of text input, thus imbuing the image with toxic semantics. Subsequently, an adversarial text suffix is integrated and co-optimized with the adversarial image prefix to maximize the probability of eliciting affirmative responses to various harmful instructions. The discovered adversarial image prefix and text suffix are collectively denoted as a Universal Master Key (UMK). When integrated into various malicious queries, UMK can circumvent the alignment defenses of VLMs and lead to the generation of objectionable content, known as jailbreaks. The experimental results demonstrate that our universal attack strategy can effectively jailbreak MiniGPT-4 with a 96% success rate, highlighting the fragility of VLMs and the exigency for new alignment strategies. Codes are available at https://github.com/roywang021/UMK. Disclaimer: This paper contains potentially disturbing and offensive content.
Face swapping, the technique of transferring the identity from one face to another, merges as a field with significant practical applications. However, previous swapping methods often result in visible artifacts. To address this issue, in our paper, we propose CodeSwap, a symmetrical framework to achieve face swapping with high-fidelity and realism. Specifically, our method firstly utilizes a codebook that captures the knowledge of high quality facial features. Building on this foundation, the face swapping is then converted into the code manipulation task in a code space. To achieve this, we design a Transformer-based architecture to update each code independently, which enable more precise manipulations. Furthermore, we incorporate a mask generator to achieve seamless blending of the generated face with the background of target image. A distinctive characteristic of our method is its symmetrical approach to processing both target and source images, simultaneously extracting information from each to improve the quality of face swapping. This symmetry also simplifies the bidirectional exchange of faces in a singular operation. Through extensive experiments on ClelebA-HQ and FF++, our method is proven to not only achieve efficient identity transfer but also substantially reduce the visible artifacts.
Voice is one of the most widely used media for information transmission in human society. While high-quality synthetic voices are extensively utilized in various applications, they pose significant risks to content security and trust building. Numerous studies have concentrated on AI-synthesized voice detection to mitigate these risks, with many claiming to achieve promising performance. However, recent research has demonstrated that fake voice detectors suffer from serious overfitting to speaker-irrelative features (SiFs) and cannot be used in real-world scenarios. In this paper, we analyze the limitations of existing fake voice detectors and propose a new design philosophy, guiding the detection model to prioritize learning human voice features rather than the difference between the human voice and the synthetic voice. Based on this philosophy, we propose a novel AI-synthesized voice detection framework named SiFSafer, which uses pre-trained speech representation models to enhance the learning of feature distribution in human voices and the adapter fine-tuning to optimize the performance. The evaluation shows that the average EERs of existing fake voice detectors in the ASVspoof datasets can exceed 20% if the SiFs like silence segments are removed, while SiFSafer achieves an EER of less than 8%, indicating that SiFSafer is robust to SiFs and strongly resistant to existing attacks.