ACM MM 2024 Papers

Open-vocabulary multi-object tracking (MOT) aims to track arbitrary objects encountered in the real world beyond the training set. However, recent methods rely solely on instance-level detection and association of novel objects, which may not consider the valuable fine-grained semantic representations of the targets within key and reference frames. In this paper, we propose a Global and Local Awareness open-vocabulary MOT method (GLATrack), which learns to tackle the task of real-world MOT from both global and instance-level perspectives. Specifically, we introduce a region-aware feature enhancement module to refine global knowledge for complementing local target information, which enhances semantic representation and bridges the distribution gap between the image feature map and the pooled regional features. We propose a bidirectional semantic complementarity strategy to mitigate semantic misalignment arising from missing target information in key frames, which dynamically selects valuable information within reference frames to enrich object representation during the knowledge distillation process. Furthermore, we introduce an appearance richness measurement module to provide appropriate representations for targets with different appearances. The proposed method gains an improvement of 6.9% in TETA and 5.6% in mAP on the large-scale TAO benchmark.

Cartoon parsing is an important task for cartoon-centric applications, which segments the body parts of cartoon images. Due to the complex appearances, abstract drawing styles, and irregular structures of cartoon characters, cartoon parsing remains a challenging task. In this paper, a novel approach, named CartoonNet, is proposed for cartoon parsing, in which semantic consistency and structure correlation are integrated to address the visual diversity and structural complexity for cartoon parsing. A memory-based semantic consistency module is designed to learn the diverse appearances exhibited by cartoon characters. The memory bank stores features of diverse samples and retrieves the samples related to new samples for consistency, which aims to improve the semantic reasoning capability of the network. A self-attention mechanism is employed to conduct consistency learning among diverse body parts belong to the retrieved samples and new samples. To capture the intricate structural information of cartoon images, a structure correlation module is proposed. Leveraging graph attention networks and a main body-aware mechanism, the proposed approach enables structural correlation, allowing it to parse cartoon images with complex structures. Experiments conducted on cartoon parsing and human parsing datasets demonstrate the effectiveness of the proposed method, which outperforms the state-of-the-art approaches for cartoon parsing and achieves competitive performance on human parsing.

Deep convolutional neural networks have made significant breakthroughs in medical image classification, under the assumption that training samples from all classes are simultaneously available. However, in real-world medical scenarios, there's a common need to continuously learn about new diseases, leading to the emerging field of class incremental learning (CIL) in the medical domain. Typically, CIL suffers from catastrophic forgetting when trained on new classes. This phenomenon is mainly caused by the imbalance between old and new classes, and it becomes even more challenging with imbalanced medical datasets. In this work, we introduce two simple yet effective plug-in methods to mitigate the adverse effects of the imbalance. First, we propose a CIL-balanced classification loss to mitigate the classifier bias toward majority classes via logit adjustment. Second, we propose a distribution margin loss that not only alleviates the inter-class overlap in embedding space but also enforces the intra-class compactness. We evaluate the effectiveness of our method with extensive experiments on three benchmark datasets (CCH5000, HAM10000, and EyePACS). The results demonstrate that our approach outperforms state-of-the-art methods.

Visual-language models based on CLIP have shown remarkable abilities in general few-shot image classification. However, their performance drops in specialized fields such as healthcare or agriculture, because CLIP's pre-training does not cover all category data. Existing methods excessively depend on the multi-modal information representation and alignment capabilities acquired from CLIP pre-training, which hinders accurate generalization to unfamiliar domains. To address this issue, this paper introduces a novel visual-language collaborative representation network (MCRNet), aiming at acquiring a generalized capability for collaborative fusion and representation of multi-modal information. Specifically, MCRNet learns to generate relational matrices from an information fusion perspective to acquire aligned multi-modal features. This relationship generation strategy is category-agnostic, so it can be generalized to new domains. A class-adaptive fine-tuning inference technique is also introduced to help MCRNet efficiently learn alignment knowledge for new categories using limited data. Additionally, the paper establishes a new broad-domain few-shot image classification benchmark containing seven evaluation datasets from five domains. Comparative experiments demonstrate that MCRNet outperforms current state-of-the-art models, achieving an average improvement of 13.06% and 13.73% in the 1-shot and 5-shot settings, highlighting the superior performance and applicability of MCRNet across various domains.

Pre-trained Vision-Language Models (VLMs) have shown great ability in various Vision-Language tasks. However, these VLMs exhibit inherent vulnerabilities to transferable adversarial examples, which could potentially undermine their performance and reliability in real-world applications. Cross-modal interactions have been demonstrated to be the key point to boosting adversarial transferability, but the utilization of them is limited in existing multimodal adversarial attacks. Stable Diffusion, which contains multiple cross-attention modules, possesses great potential in facilitating adversarial transferability by leveraging abundant cross-modal interactions. Therefore, We propose a Multimodal Diffusion-based Attack (MDA), which conducts adversarial attacks against VLMs using Stable Diffusion. Specifically, MDA initially generates adversarial text, which is subsequently utilized to optimize the adversarial image during the diffusion process. Besides leveraging adversarial text in calculating downstream loss, MDA also takes it as the guiding prompt in adversarial image generation during the denoising process, which enriches the ways of cross-modal interactions, thus strengthening the adversarial transferability. Compared with pixel-based attacks, MDA introduces perturbations in the latent space rather than pixel space to manipulate high-level semantics, which is also beneficial to improving adversarial transferability. Experimental results demonstrate that the adversarial examples generated by MDA are highly transferable across different VLMs on different downstream tasks, surpassing state-of-the-art methods by a large margin.

Singing Voice Synthesis (SVS) has significantly advanced with deep generative models, achieving high audio quality but still struggling with musicality, mainly due to the lack of performance control over timing, dynamics, and pitch, which are essential for music expression. Additionally, integrating data and supporting diverse languages and styles in SVS remain challenging. To tackle these issues, this paper presents ExpressiveSinger, an SVS framework that leverages a cascade of diffusion models to generate realistic singing across multiple languages, styles, and techniques from scores and lyrics. Our approach begins with consolidating, cleaning, annotating, and processing public singing datasets, developing a multilingual phoneme set, and incorporating different musical styles and techniques. We then design methods for generating expressive performance control signals including phoneme timing, F0 curves, and amplitude envelopes, which enhance musicality and model consistency, introduce more controllability, and reduce data requirements. Finally, we generate mel-spectrograms and audio from performance control signals with style guidance and singer timbre embedding. Our models also enable trained singers to sing in new languages and styles. Several listening tests reveal both musicality and controllability of our generated singing compared with existing works and human singing. We release the data for future research. Demo: https://shuqid.net/expressive-singing-synthesis.

Spiking Neural Networks (SNNs) have indeed shown remarkable promise in the field of computer vision, emerging as a low-energy alternative to traditional Artificial Neural Networks (ANNs). However, SNNs also face several challenges: i) Existing SNNs are not purely additive and involve a substantial amount of floating-point computations, which contradicts the original design intention of adapting to neuromorphic chips; ii) The incorrect positioning of convolutional and pooling layers relative to spiking layers leads to reduced accuracy; iii) Leaky Integrate-and-Fire (LIF) neurons have limited capability in representing local information, which is disadvantageous for downstream visual tasks like semantic segmentation.

Face Recognition (FR) systems can be easily deceived by adversarial examples that manipulate benign face images through imperceptible perturbations. Adversarial attacks on FR encompass two types: impersonation (targeted) attacks and dodging (untargeted) attacks. Previous methods often achieve a successful impersonation attack on FR, however, it does not necessarily guarantee a successful dodging attack on FR in the black-box setting. In this paper, our key insight is that the generation of adversarial examples should perform both impersonation and dodging attacks simultaneously. To this end, we propose a novel attack method termed as Adversarial Pruning (Adv-Pruning), to fine-tune existing adversarial examples to enhance their dodging capabilities while preserving their impersonation capabilities. Adv-Pruning consists of Priming, Pruning, and Restoration stages. Concretely, we propose Adversarial Priority Quantification to measure the region-wise priority of original adversarial perturbations, identifying and releasing those with minimal impact on absolute model output variances. Then, Biased Gradient Adaptation is presented to adapt the adversarial examples to traverse the decision boundaries of both the attacker and victim by adding perturbations favoring dodging attacks on the vacated regions, preserving the prioritized features of the original perturbations while boosting dodging performance. As a result, we can maintain the impersonation capabilities of original adversarial examples while effectively enhancing dodging capabilities. Comprehensive experiments demonstrate the superiority of our method compared with state-of-the-art adversarial attack methods.

Fine-grained leaf image retrieval (FGLIR) is a new unsupervised pattern recognition task in content-based image retrieval (CBIR). It aims to distinguish varieties/cultivars of leaf images within a certain plant species and is more challenging than general leaf image retrieval task due to the inherently subtle differences across different cultivars. In this study, we for the first time investigate the possible way to mine the spatial structure and contextual information from the activation of the convolutional layers of CNN networks for FGLIR. For achieving this goal, we design a novel geometrical structure, named Triplet Patch-Pairs Composite Structure (TPCS), consisting of three symmetric patch pairs segmented from the leaf images in different orientations. We extract CNN feature map for each patch in TPCS and measure the difference between the feature maps of the patch pair for constructing local deep self-similarity descriptor. By varying the size of the TPCS, we can yield multi-scale deep self-similarity descriptors. The final aggregated local deep self-similarity descriptors, named Structural Deep Patch Representation (SDePR), not only encode the spatial structure and contextual information of leaf images in deep feature domain, but also are invariant to the geometrical transformations. The extensive experiments of applying our SDePR method to the public challenging FGLIR tasks show that our method outperforms the state-of-the-art handcrafted visual features and deep retrieval models.

Achieving 3D hand-object pose estimation in interaction scenarios is challenging due to the severe occlusion generated during the interaction. Existing methods address this issue by utilizing the correlation between the hand and object poses as additional cues. They usually first extract the hand and object features from their respective regions and then refine them with each other. However, this paradigm disregards the role of a broad range of image context. To address this problem, we propose a novel and robust approach that learns a broad range of context by imposing priors. First, we build this approach using stacked transformer decoder layers. These layers are required for extracting image-wide context and regional hand or object features by constraining cross-attention operations. We share the context decoder layer parameters between the hand and object pose estimations to avoid interference in the context-learning process. This imposes a prior, indicating that the hand and object are mutually the most important context for each other, significantly enhancing the robustness of obtained context features. Second, since they play different roles, we provide customized feature maps for the context, hand, and object decoder layers. This strategy facilitates the disentanglement of these layers, reducing the feature learning complexity. Finally, we conduct extensive experiments on the popular HO3D and Dex-YCB databases. The experimental results indicate that our method significantly outperforms state-of-the-art approaches and can be applied to other hand pose estimation tasks. Code is available at https://github.com/zskuang58/LCP.

Facial movements play a crucial role in conveying altitude and intentions, and facial optical flow provides a dynamic and detailed representation of it. However, the scarcity of datasets and a modern baseline hinders the progress in facial optical flow research. This paper proposes FacialFlowNet (FFN), a novel large-scale facial optical flow dataset, and the Decomposed Facial Flow Model (DecFlow), the first method capable of decomposing facial flow. FFN comprises 9,635 identities and 105,970 image pairs, offering unprecedented diversity for detailed facial and head motion analysis. DecFlow features a facial semantic-aware encoder and a decomposed flow decoder, excelling in accurately estimating and decomposing facial flow into head and expression components. Comprehensive experiments demonstrate that FFN significantly enhances the accuracy of facial flow estimation across various optical flow methods, achieving up to an 11% reduction in Endpoint Error (EPE) (from 3.91 to 3.48). Moreover, DecFlow, when coupled with FFN, outperforms existing methods in both synthetic and real-world scenarios, enhancing facial expression analysis. The decomposed expression flow achieves a substantial accuracy improvement of 18% (from 69.1% to 82.1%) in micro-expressions recognition. These contributions represent a significant advancement in facial motion analysis and optical flow estimation. Codes and datasets can be found.

Homography estimation is the task of determining the transformation from an image pair. Our approach focuses on employing detector-free feature matching methods to address this issue. Previous work has underscored the importance of incorporating semantic information, however there still lacks an efficient way to utilize semantic information. Previous methods suffer from treating the semantics as a pre-processing, causing the utilization of semantics overly coarse-grained and lack adaptability when dealing with different tasks. In our work, we seek another way to use the semantic information, that is semantic-aware feature representation learning framework. Based on this, we propose SRMatcher, a new detector-free feature matching method, which encourages the network to learn integrated semantic feature representation. Specifically, to capture precise and rich semantics, we leverage the capabilities of recently popularized vision foundation models (VFMs) trained on extensive datasets. Then, a cross-images Semantic-aware Fusion Block (SFB) is proposed to integrate its fine-grained semantic features into the feature representation space. In this way, by reducing errors stemming from semantic inconsistencies in matching pairs, our proposed SRMatcher is able to deliver more accurate and realistic outcomes. Extensive experiments show that SRMatcher surpasses solid baselines and attains SOTA results on multiple real-world datasets. Compared to the previous SOTA approach GeoFormer, SRMatcher increases the area under the cumulative curve (AUC) by about 11% on HPatches. Additionally, the SRMatcher could serve as a plug-and-play framework for other matching methods like LoFTR, yielding substantial precision improvement.

The field of floorplan generation has attracted significant interest from the community. Remarkably, recent advances in generative models have markedly enhanced the development of this field. However, generating floorplans that satisfy various conditions remains a challenging task. This paper proposes a learning framework, named Cons2Plan, for automatically and high-quality generating vector floorplans from various conditions. The input conditions can be graphs, boundaries, or a combination of both. The conditional diffusion model is the core component of our Cons2Plan. The denoising network uses a conditional embedding module to incorporate the conditions during the reverse process. Additionally, Cons2Plan incorporates a two-stage approach that generates graph conditions based on boundaries. It uses three networks for node prediction and a novel conditional edge generation diffusion model, named CEDM, for edge generation. We conduct qualitative evaluations, quantitative comparisons, and ablation studies to show that our method produces better floorplans than state-of-the-art methods.

Emotion recognition based on electroencephalogram (EEG) has garnered increasing attention in recent years due to non-invasiveness and high reliability of EEG measurements. Despite the promising performance achieved by numerous existing methods, several challenges persist. Firstly, there is the challenge of emotional label noise, stemming from the assumption that emotions remain consistently evoked and stable throughout the entirety of video observation. Such an assumption proves difficult to uphold in practical experimental settings, leading to discrepancies between EEG signals and anticipated emotional states. In addition, there's a need for comprehensive capture of temporal-spatial-spectral characteristics of EEG signals and cope with low signal-to-noise ratio (SNR) issues. To tackle these challenges, we propose a comprehensive pipeline named REmoNet, which leverages novel self-supervised techniques and multi-regularized co-learning. Two self-supervised methods, including masked channel modeling via temporal-spectral transformation and emotion contrastive learning, are introduced to facilitate the comprehensive understanding and extraction of emotion-relevant EEG representations during pre-training. Additionally, fine-tuning with multi-regularized co-learning exploits feature-dependent information through intrinsic similarity, resulting in mitigating emotional label noise. Experimental evaluations on two public datasets demonstrate that our proposed approach, REmoNet, surpasses existing state-of-the-art methods, showcasing its effectiveness in simultaneously addressing raw EEG signals and noisy emotional labels.

Skeleton-based zero-shot action recognition aims to recognize unknown human actions based on the learned priors of the known skeleton-based actions and a semantic descriptor space shared by both known and unknown categories. However, previous works mainly focus on establishing the bridges between the known skeleton representation space and semantic descriptions space at the coarse-grained level for recognizing unknown action categories, ignoring the fine-grained alignment of these two spaces, resulting in suboptimal performance in distinguishing high-similarity action categories. To address these challenges, we propose a novel method via Side information and dual-prompTs learning for skeleton-based zero-shot Action Recognition (STAR) at the fine-grained level. Specifically, 1) we decompose the skeleton into several parts based on its topology structure and introduce the side information concerning multi-part descriptions of human body movements for alignment between the skeleton and the semantic space at the fine-grained level; 2) we design the visual-attribute and semantic-part prompts to improve the intra-class compactness within the skeleton space and inter-class separability within the semantic space, respectively, to distinguish the high-similarity actions. Extensive experiments show that our method achieves state-of-the-art performance in ZSL and GZSL settings on NTU RGB+D, NTU RGB+D 120, and PKU-MMD datasets.

Deep network-based image Compressive Sensing (CS) has attracted much attention in recent years. However, there still exist the following two issues: 1) Existing methods typically use fixed-scale sampling, which leads to limited insights into the image content. 2) Most pre-trained models can only handle fixed sampling rates and fixed block scales, which restricts the scalability of the model. In this paper, we propose a novel scale-aware scalable CS network (dubbed S2-CSNet), which achieves scale-aware adaptive sampling, fine granular scalability and high-quality reconstruction with one single model. Specifically, to enhance the scalability of the model, a structural sampling matrix with a predefined order is first designed, which is a universal sampling matrix that can sample multi-scale image blocks with arbitrary sampling rates. Then, based on the universal sampling matrix, a distortion-guided scale-aware scheme is presented to achieve scale-variable adaptive sampling, which predicts the reconstruction distortion at different sampling scales from the measurements and select the optimal division scale for sampling. Furthermore, a multi-scale hierarchical sub-network under a well-defined compact framework is put forward to reconstruct the image. In the multi-scale feature domain of the sub-network, a dual spatial attention is developed to explore the local and global affinities between dense feature representations for deep fusion. Extensive experiments manifest that the proposed S2-CSNet outperforms existing state-of-the-art CS methods.

3D occupancy prediction (OCC) aims to estimate and predict the semantic occupancy state of the surrounding environment, which is crucial for scene understanding and reconstruction in the real world. However, existing methods for 3D OCC mainly rely on surround-view camera images, whose performance is still insufficient in some challenging scenarios, such as low-light conditions. To this end, we propose a new multi-modal fusion network for 3D occupancy prediction by fusing features of LiDAR point clouds and surround-view images, called FusionOcc. Our model fuses features of these two modals in 2D and 3D space, respectively. By integrating the depth information from point clouds, a cross-modal fusion module is designed to predict a 2D dense depth map, enabling an accurate depth estimation and a better transition of 2D image features into 3D space. In addition, features of voxelized point clouds are aligned and merged with image features converted by a view-transformer in 3D space. Experiments show that FusionOcc establishes the new state of the art on Occ3D-nuScenes dataset, achieving a mIoU score of 35.94% (without visibility mask) and 56.62% (with visibility mask), showing an average improvement of 3.42% compared to the best previous method. Our work provides a new baseline for further research in multi-modal fusion for 3D occupancy prediction. Codes will be made publicly at https://github.com/ShuoZhang-code/FusionOcc.