Diffusion models have exhibited impressive prowess in the text-to-image task. Recent methods add image-level structure controls, e.g., edge and depth maps, to manipulate the generation process together with text prompts to obtain desired images. This controlling process is globally operated on the entire image, which limits the flexibility of control regions. In this paper, we explore a novel and practical task setting: local control. It focuses on controlling specific local region according to user-defined image conditions, while the remaining regions are only conditioned by the original text prompt. However, it is non-trivial to achieve it. The naive manner of directly adding local conditions may lead to the local control dominance problem, which forces the model to focus on the controlled region and neglect object generation in other regions. To mitigate this problem, we propose Regional Discriminate Loss to update the noised latents, aiming at enhanced object generation in non-control regions. Furthermore, the proposed Focused Token Response suppresses weaker attention scores which lack the strongest response to enhance object distinction and reduce duplication. Lastly, we adopt Feature Mask Constraint to reduce quality degradation in images caused by information differences across the local control region. All proposed strategies are operated at the inference stage. Extensive experiments demonstrate that our method can synthesize high-quality images aligned with the text prompt under local control conditions.
论文检索
输入标题、作者或关键词，从 11,272 篇学术成果中精准定位
The Audio-Visual Question Answering (AVQA) task involves extracting question-related audio-visual clues from both temporal and spatial perspectives to answer questions accurately. Despite the promising performance of existing multi-modal AVQA models, thanks to large-scale pre-trained models, challenges remain in the field. Firstly, aligning audio-visual information across temporal and spatial dimensions is difficult. Secondly, the fusion of audio-visual information is often weighted inadequately, limiting model performance. To address the above issues, we design the Audio-Visual Adaptive Fusion Network (AVAF-Net), which uses contrastive learning to align audio-visual information temporally and spatially and adaptively adjusts fusion weights based on the question. Specifically, we initially align visual and audio information temporally through a temporal-alignment contrastive loss. This is followed by an audio-visual clue-mining module that highlights question-related cues, aligning them with the vocal region spatially using spatial alignment contrastive loss. Additionally, a question-oriented adaptive fusion module assigns different weights to audio and visual modalities based on the question content and then fuses them. The fused audio-visual cues are finally used to predict the answer. Extensive experiments on the MUSIC-AVQA dataset show that AVAF-Net surpasses all baseline models, with a maximum improvement of 15.90% in average accuracy and an average improvement of 9.80%.
Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching
PDF ↗Open-vocabulary semantic segmentation (OVSS) aims to segment images of arbitrary categories specified by class labels. While previous approaches relied on extensive image-text pairs or dense semantic annotations, recent training-free methods attempted to overcome these limitations by constructing semantic prototypes in the construction stage and image-to-image matching (i.e., prototype matching) during testing. However, these methods often struggle to effectively capture the visual characteristics of categories and fail to utilize local features during prototype matching. To deal with these problems, we propose a novel training-free framework for OVSS that constructs diverse prototypes and performs fine-grained sub-region matching. Specifically, our method leverages Large Language Models (LLMs) to guide support image generation by descriptions of different attributes of categories and employs coarse-fine clustering to obtain diverse and robust part-level prototypes in the construction stage. During testing, we propose a sub-region matching method, which assigns part-level prototypes to sub-regions utilizing optimal transport, to fully utilize local image features among part-level prototypes. Extensive experiments demonstrate the effectiveness of our method and show that our method achieves state-of-the-art performance, outperforming previous methods across five datasets.
Deep hashing has been widely used for large-scale approximate nearest neighbor search due to its storage and search efficiency. However, existing deep hashing methods predominantly rely on abundant training data, leaving the more challenging scenario of low-resource adaptation for deep hashing relatively underexplored. This setting involves adapting pre-trained models to downstream tasks with only an extremely small number of training samples available. Our preliminary benchmarks reveal that current methods suffer significant performance degradation due to the distribution shift caused by limited training samples. To address these challenges, we introduce Class-Calibration LoRA (CLoRA), a novel plug-and-play approach that dynamically constructs low-rank adaptation matrices by leveraging class-level textual knowledge embeddings. CLoRA effectively incorporates prior class knowledge as anchors, enabling parameter-efficient fine-tuning while maintaining the original data distribution. Furthermore, we propose Knowledge-Guided Discrete Optimization (KIDDO), a framework to utilize class knowledge to compensate for the scarcity of visual information and enhance the discriminability of hash codes. Extensive experiments demonstrate that our proposed method, Knowledge- Anchored Low-Resource Adaptation Hashing (KALAHash), significantly boosts retrieval performance and achieves a 4× data efficiency in low-resource scenarios.
Open vocabulary semantic segmentation is a hot topic in research, focusing on segmenting and recognizing a diverse array of categories in varied environments, including those previously unknown, thereby holding significant practical value. Mainstream studies utilize the CLIP model for direct semantic segmentation (denoted as “forward methods”), which often struggles to represent underrepresented categories effectively. To address this issue, this paper introduces a novel approach Excluding the ImpossibLe Semantic Segmentation Network (ELSE-Net) based on reverse thinking. By excluding improbable categories, ELSE-Net narrows the selection range for forward methods, significantly reducing the risk of misclassification. In implementation, we initially draw on leading research to design the General Processing Block (GP-Block), which generates inclusion probabilities (the likelihood of belonging to a category) by using the CLIP model cooperated with a Mask Proposal Network (MPN). We then present the EXcluding the ImPossible Block (EXP-Block), which computes exclusion probabilities (the likelihood of not belonging to a category) through the CLIPN model and a custom-designed Reverse Retrieval Adapter (R2-Adapter). These exclusion probabilities are subsequently used to refine the inclusion probabilities, which are ultimately employed to annotate class-agnostic masks. Moreover, the core component of our EXP-Block is model-agnostic, enabling it to enhance the capabilities of existing frameworks. Experimental results from four benchmark datasets validate the effectiveness of ELSE-Net and underscore the seamless model-agnostic functionality of the EXP-Block.
The Audio-Visual Video Parsing task aims to recognize and temporally localize all events occurring in either the audio or visual stream, or both. Capturing accurate event semantics for each audio/visual segment is vital. Prior works directly utilize the extracted holistic audio and visual features for intra- and cross-modal temporal interactions. However, each segment may contain multiple events, resulting in semantically mixed holistic features that can lead to semantic interference during intra- or cross-modal interactions: the event semantics of one segment may incorporate semantics of unrelated events from other segments. To address this issue, our method begins with a Class-Aware Feature Decoupling (CAFD) module, which explicitly decouples the semantically mixed features into distinct class-wise features, including multiple event-specific features and a dedicated background feature. The decoupled class-wise features enable our model to selectively aggregate useful semantics for each segment from clearly matched classes contained in other segments, preventing semantic interference from irrelevant classes. Specifically, we further design a Fine-Grained Semantic Enhancement module for encoding intra- and cross-modal relations. It comprises a Segment-wise Event Co-occurrence Modeling (SECM) block and a Local-Global Semantic Fusion (LGSF) block. The SECM exploits inter-class dependencies of concurrent events within the same timestamp with the aid of a novel event co-occurrence loss. The LGSF further enhances the event semantics of each segment by incorporating relevant semantics from more informative global video features. Extensive experiments validate the effectiveness of the proposed modules and loss functions, resulting in a new state-of-the-art parsing performance.
Large-scale text-to-image diffusion models, (e.g., DALL-E, SDXL) are capable of generating famous persons by simply referring to their names. Is it possible to make such models generate generic identities as simple as the famous ones, e.g., just use a name? In this paper, we explore the existence of a ``Name Space'', where any point in the space corresponds to a specific identity. Fortunately, we find some clues in the feature space spanned by text embedding of celebrities' names. Specifically, we first extract the embeddings of celebrities' names in the Laion5B dataset with the text encoder of diffusion models. Such embeddings are used as supervision to learn an encoder that can predict the name (actually an embedding) of a given face image. We experimentally find that such name embeddings work well in promising the generated image with good identity consistency. Note that like the names of celebrities, our predicted name embeddings are disentangled from the semantics of text inputs, making the original generation capability of text-to-image models well-preserved. Moreover, by simply plugging such name embeddings, all variants (e.g., from Civitai) derived from the same base model (i.e., SDXL) readily become identity-aware text-to-image models.
Adaptive Wavelet-Positional Encoding for High-Frequency Information Learning in Implicit Neural Representation
PDF ↗Implicit Neural Representation (INR) has shown great potential in constructing the complex nature signal as a continuous implicit function. However, the representation results are incomplete since different components of the signal correspond to different frequencies and neural network inherently tends to low-frequency convergence. In this paper, we propose the adaptive Wavelet-Positional Encoding (WPE) to precisely represent content under different frequency distributions for coordinate-based implicit representations. The High-Frequency Perception (HFP) method is first proposed to query locations of high-frequency components from input signals, which can be indicated as local centers of WPE. Then, motivated by wavelet series regression, we present to embed these queried low-dimensional coordinate inputs into wavelet-frequency space by WPE to represent fine details of target signals. Experiments demonstrate that the proposed method can be integrated into various INR methods without modifying training frameworks while significantly improving their performance in 1D signal fitting, 2D image regression, and even 3D scene representation.
In recent years, applying multi-modal large language models (MLLMs) in various fields has achieved remarkable success. However, as the foundation model for many downstream tasks, MLLMs comprise the well-known Transformer network, which has a less efficient quadratic computation complexity. In this study, we introduce Cobra, a multi-modal large-scale language model built upon a state-space model, which has demonstrated significant potential in efficiently handling long sequences with fast inference and linear scalability concerning sequence length. Specifically, Cobra involves replacing Transformer-based backbone models (e.g., LLaMA or Phi) with pre-trained Mamba language models. We then empirically explore effective strategies for aligning visual and textual modalities and integrating various pre-trained Mamba model variants with visual encoders. Experiments across various multi-modal benchmarks demonstrate that: (i) Cobra performs 3× ∼ 4× faster than the most computationally efficient state-of-the-art methods, e.g., LLaVA-Phi and MobileVLM v2. Additionally, its performance is significantly enhanced thanks to the implementation of linear sequential modeling. (ii) Cobra fine-tunes a small parameter (∼48% of model parameters), leading to a significant improvement in overall performance compared to LLaVA.
World models have demonstrated superiority in autonomous driving, particularly in the generation of multi-view driving videos. However, significant challenges still exist in generating customized driving videos. In this paper, we propose DriveDreamer-2, which incorporates a Large Language Model (LLM) to facilitate the creation of user-defined driving videos. Specifically, a trajectory generation function library is developed to produce trajectories that conform to user descriptions. Subsequently, an HDMap generator is designed to learn the mapping from trajectories to road structures. Ultimately, we propose the Unified Multi-View Model (UniMVM) to enhance temporal and spatial coherence in the generated multi-view driving videos. To the best of our knowledge, DriveDreamer-2 is the first world model to generate customized driving videos, and it can generate uncommon driving videos (e.g., vehicles abruptly cut in) in a user-friendly manner. Besides, experimental results demonstrate that the generated videos enhance the training of driving perception methods (e.g., 3D detection and tracking). Furthermore, video generation quality of DriveDreamer-2 surpasses other state-of-the-art methods, showcasing FID and FVD scores of 11.2 and 55.7, representing relative improvements of ~30% and ~50%.
Due to the density inconsistency and distribution difference between cross-source point clouds, previous methods fail in cross-source point cloud registration. We propose a density-robust feature extraction and matching scheme to achieve robust and accurate cross-source registration. To address the density inconsistency between cross-source data, we introduce a density-robust encoder for extracting density-robust features. To tackle the issue of challenging feature matching and few correct correspondences, we adopt a loose-to-strict matching pipeline with a ``loose generation, strict selection'' idea. Under it, we employ a one-to-many strategy to loosely generate initial correspondences. Subsequently, high-quality correspondences are strictly selected to achieve robust registration through sparse matching and dense matching. On the challenging Kinect-LiDAR scene in the cross-source 3DCSR dataset, our method improves feature matching recall by 63.5 percentage points (pp) and registration recall by 57.6 pp. It also achieves the best performance on 3DMatch, while maintaining robustness under diverse downsampling densities.
Hand-object interaction modeling from a single RGB image is a significantly challenging task. Previous works typically reconstruct hand-object interactions as texture-less meshes, ignoring photo-realistic image generation. In this work, we introduce the HO123, a novel method to synthesize novel-view hand-object interaction images from a single image. To this end, we first train a 2D diffusion prior. Given the camera pose in novel views, our approach transfers the camera information into explicit hand representations, including hand depth and skeleton images. We propose a global hand embedding to control the diffusion model based on these hand representations. We then learn a 3D Gaussian splatting for novel-view rendering using the diffusion prior. However, occluded objects present a persistent challenge. To address this issue, we further introduce local hand embedding, where a contact field is defined in the 3D Gaussian Splatting. We leverage contact information to guide the rendering in the contact field. Extensive experiments on the HO3D and DexYCB datasets demonstrate that our method significantly outperforms state-of-the-art novel-view synthesis for hand-object interactions.
Multi-scale Activation, Selection, and Aggregation: Exploring Diverse Cues for Fine-Grained Bird Recognition
PDF ↗Given the critical role of birds in ecosystems, Fine-Grained Bird Recognition (FGBR) has gained increasing attention, particularly in distinguishing birds within similar subcategories. Although Vision Transformer (ViT)-based methods often outperform Convolutional Neural Network (CNN)-based methods in FGBR, recent studies reveal that the limited receptive field of plain ViT model hinders representational richness and makes them vulnerable to scale variance. Thus, enhancing the multi-scale capabilities of existing ViT-based models to overcome this bottleneck in FGBR is a worthwhile pursuit. In this paper, we propose a novel framework for FGBR, namely Multi-scale Diverse Cues Modeling (MDCM), which explores diverse cues at different scales across various stages of a multi-scale Vision Transformer (MS-ViT) in an ``Activation-Selection-Aggregation'' paradigm. Specifically, we first propose a multi-scale cue activation module to ensure the discriminative cues learned at different stage are mutually different. Subsequently, a multi-scale token selection mechanism is proposed to remove redundant noise and highlight discriminative, scale-specific cues at each stage. Finally, the selected tokens from each stage are independently utilized for bird recognition, and the recognition results from multiple stages are adaptively fused through a multi-scale dynamic aggregation mechanism for final model decisions. Both qualitative and quantitative results demonstrate the effectiveness of our proposed MDCM, which outperforms CNN- and ViT-based models on several widely-used FGBR benchmarks.
Image manipulation localization (IML) is a critical technique in media forensics, focusing on identifying tampered regions within manipulated images. Most existing IML methods require extensive training on labeled datasets with both image-level and pixel-level annotations. These methods often struggle with new manipulation types and exhibit low generalizability. In this work, we propose a training-free IML approach using diffusion models. Our method adaptively selects an appropriate number of diffusion timesteps for each input image in the forward process and performs both conditional and unconditional reconstructions in the backward process without relying on external conditions. By comparing these reconstructions, we generate a localization map highlighting regions of manipulation based on inconsistencies. Extensive experiments were conducted using sixteen state-of-the-art (SoTA) methods across six IML datasets. The results demonstrate that our training-free method outperforms SoTA unsupervised and weakly-supervised techniques. Furthermore, our method competes effectively against fully-supervised methods on novel (unseen) manipulation types.
Recent approaches to VO have significantly improved performance by using deep networks to predict optical flow between video frames. However, existing methods still suffer from noisy and inconsistent flow matching, making it difficult to handle challenging scenarios and long-sequence estimation.To overcome these challenges, we introduce Spatio-Temporal Visual Odometry (STVO), a novel deep network architecture that effectively leverages inherent spatio-temporal cues to enhance the accuracy and consistency of multi-frame flow matching. With more accurate and consistent flow matching, STVO can achieve better pose estimation through the bundle adjustment (BA).Specifically, STVO introduces two innovative components: 1) the Temporal Propagation Module that utilizes multi-frame information to extract and propagate temporal cues across adjacent frames, maintaining temporal consistency; 2) the Spatial Activation Module that utilizes geometric priors from the depth maps to enhance spatial consistency while filtering out excessive noise and incorrect matches.Our STVO achieves state-of-the-art performance on TUM-RGBD, EuRoc MAV, ETH3D and KITTI Odometry benchmarks. Notably, it improves accuracy by 77.8% on ETH3D benchmark and 38.9% on KITTI Odometry benchmark over the previous best methods.
Zero-Shot Learning in Industrial Scenarios: New Large-Scale Benchmark, Challenges and Baseline
PDF ↗Large Visual Language Models (LVLMs) have achieved remarkable success in vision tasks. However, the significant differences between industrial and natural scenes make applying LVLMs challenging. Existing LVLMs rely on user-provided prompts to segment objects. This often leads to suboptimal performance due to the inclusion of irrelevant pixels. In addition, the scarcity of data also makes the application of LVLMs in industrial scenarios remain unexplored. To fill this gap, this paper proposes an open industrial dataset and a Refined Text-Visual Prompt (RTVP) for zero-shot industrial defect detection. First, this paper constructs the Multi-Modal Industrial Open Dataset (MMIO) containing 80K+ samples. MMIO contains diverse industrial categories, including 6 super categories and 18 subcategories. MMIO is the first large-scale multi-scenes pre-training dataset for industrial zero-shot learning, and provides valuable training data for open models in future industrial scenarios. Based on MMIO, this paper provides a RTVP specifically for industrial zero-shot tasks. RTVP has two significant advantages: First, this paper designs an expert-guided large model domain adaptation mechanism and designs an industrial zero-shot method based on Mobile-SAM, which enhances the generalization ability of large models in industrial scenarios. Second, RTVP automatically generates visual prompts directly from images and considers text-visual prompt interactions ignored by previous LVLM, improving visual and textual content understanding. RTVP achieves SOTA with 42.2% and 24.7% AP in zero-shot and closed scenes of MMIO.
Current hair transfer methods struggle to handle diverse and intricate hairstyles, limiting their applicability in real-world scenarios. In this paper, we propose a novel diffusion-based hair transfer framework, named Stable-Hair, which robustly transfers a wide range of real-world hairstyles to user-provided faces for virtual hair try-on. To achieve this goal, our Stable-Hair framework is designed as a two-stage pipeline. In the first stage, we train a Bald Converter alongside stable diffusion to remove hair from the user-provided face images, resulting in bald images. In the second stage, we specifically designed a Hair Extractor and a Latent IdentityNet to transfer the target hairstyle with highly detailed and high-fidelity to the bald image. The Hair Extractor is trained to encode reference images with the desired hairstyles, while the Latent IdentityNet ensures consistency in identity and background. To minimize color deviations between source images and transfer results, we introduce a novel Latent ControlNet architecture, which functions as both the Bald Converter and Latent IdentityNet. After training on our curated triplet dataset, our method accurately transfers highly detailed and high-fidelity hairstyles to the source images. Extensive experiments demonstrate that our approach achieves state-of-the-art performance compared to existing hair transfer methods.
Adversarial attack and defense have been extensively explored in classification tasks, but their study in semantic segmentation remains limited. Moreover, current attacks fail to act as strong underlying attacks for adversarial training (AT), making it difficult to achieve segmentation robustness against strong attacks. In this paper, we present RP-PGD, a novel Region-and-Prototype based Projected Gradient Descent attack tailored to fool segmentation models. In particular, we propose a region-based attack, which leverages a spatial-temporal way to separate the pixels into three disjoint regions, and highlights the attack on the crucial True Region and Boundary Region. Moreover, we introduce a prototype-based attack to disrupt the feature space, further enhancing the attack capability. To boost the robustness of segmentation models, we inject adversaries generated by RP-PGD into the clean data and perform AT. Extensive experiments on multiple datasets showcase that RP-PGD generates adversaries with faster convergence and stronger attack effectiveness, surpassing state-of-the-art attacks by a large margin. Consequently, RP-PGD serves as a strong underlying attack for segmentation models to perform AT, assisting them in defending against a variety of strong attacks without incurring additional computational costs during inference.
Learning representations from numerous 2D image data has shown promising performance, yet very few works apply this representations to point cloud registration. In this paper, we explore how to leverage the 2D information to assist the point cloud registration, and propose IAPReg, an Image-Assisted Partial 3D point cloud Registration framework with the multi-view images generated by the input point cloud. It is expected to enrich 3D information with 2D knowledge, and leverage 2D knowledge to assist with point cloud registration. Specifically, we create multi-view depth maps by projecting the input point cloud from several specific views, and then extract 2D and 3D features using some well-established models. To fuse the information learned from 2D and 3D modalities, inter-modality multi-view learning module is proposed to enhance geometric information and complement semantic information. Weighted SVD is a common method to reduce the impact of inaccurate correspondences on registration. However, determining the correspondence weights is not trivial. Therefore, we design a 2D-weighted SVD method, where the 2D knowledge is employed to provide weight information of correspondences. Extensive experiments perform that our method outperform the state-of-the-art method without additional 2D training data.
Generating high-quality whole-body human object interaction motion sequences is becoming increasingly important in various fields such as animation, VR/AR, and robotics. The main challenge of this task lies in determining the level of involvement of each hand given the complex shapes of objects in different sizes and their different motion trajectories, while ensuring strong grasping realism and guaranteeing the coordination of movement in all body parts. Contrasting with existing work, which either generates human interaction motion sequences without detailed hand grasping poses or only models a static grasping pose, we propose a simple yet effective framework that jointly models the relationship between the body, hands, and the given object motion sequences within a single diffusion model. To guide our network in perceiving the object's spatial position and learning more natural grasping poses, we introduce novel contact-aware losses and incorporate a data-driven, carefully designed guidance. Experimental results demonstrate that our approach outperforms the state-of-the-art method and generates plausible results.