Multimodal hate detection, which aims to identify harmful content online such as memes, is crucial for building a wholesome internet environment. Previous work has made enlightening exploration in detecting explicit hate remarks. However, most of their approaches neglect the analysis of implicit harm, which is particularly challenging as explicit text markers and demographic visual cues are often twisted or missing. The leveraged cross-modal attention mechanisms also suffer from the distributional modality gap and lack logical interpretability. To address these semantic gaps issues, we propose TOT: a topology-aware optimal transport framework to decipher the implicit harm in memes scenario, which formulates the cross-modal aligning problem as solutions for optimal transportation plans. Specifically, we leverage an optimal transport kernel method to capture complementary information from multiple modalities. The kernel embedding provides a non-linear transformation ability to reproduce a kernel Hilbert space (RKHS), which reflects significance for eliminating the distributional modality gap. Moreover, we perceive the topology information based on aligned representations to conduct bipartite graph path reasoning. The newly achieved state-of-the-art performance on two publicly available benchmark datasets, together with further visual analysis, demonstrate the superiority of TOT in capturing implicit cross-modal alignment.
相關內容
Temporal modeling is crucial for multi-frame human pose estimation. Most existing methods directly employ optical flow or deformable convolution to predict full-spectrum motion fields, which might incur numerous irrelevant cues, such as a nearby person or background. Without further efforts to excavate meaningful motion priors, their results are suboptimal, especially in complicated spatiotemporal interactions. On the other hand, the temporal difference has the ability to encode representative motion information which can potentially be valuable for pose estimation but has not been fully exploited. In this paper, we present a novel multi-frame human pose estimation framework, which employs temporal differences across frames to model dynamic contexts and engages mutual information objectively to facilitate useful motion information disentanglement. To be specific, we design a multi-stage Temporal Difference Encoder that performs incremental cascaded learning conditioned on multi-stage feature difference sequences to derive informative motion representation. We further propose a Representation Disentanglement module from the mutual information perspective, which can grasp discriminative task-relevant motion signals by explicitly defining useful and noisy constituents of the raw motion features and minimizing their mutual information. These place us to rank No.1 in the Crowd Pose Estimation in Complex Events Challenge on benchmark dataset HiEve, and achieve state-of-the-art performance on three benchmarks PoseTrack2017, PoseTrack2018, and PoseTrack21.
We present a unified framework for camera-space 3D hand pose estimation from a single RGB image based on 3D implicit representation. As opposed to recent works, most of which first adopt holistic or pixel-level dense regression to obtain relative 3D hand pose and then follow with complex second-stage operations for 3D global root or scale recovery, we propose a novel unified 3D dense regression scheme to estimate camera-space 3D hand pose via dense 3D point-wise voting in camera frustum. Through direct dense modeling in 3D domain inspired by Pixel-aligned Implicit Functions for 3D detailed reconstruction, our proposed Neural Voting Field (NVF) fully models 3D dense local evidence and hand global geometry, helping to alleviate common 2D-to-3D ambiguities. Specifically, for a 3D query point in camera frustum and its pixel-aligned image feature, NVF, represented by a Multi-Layer Perceptron, regresses: (i) its signed distance to the hand surface; (ii) a set of 4D offset vectors (1D voting weight and 3D directional vector to each hand joint). Following a vote-casting scheme, 4D offset vectors from near-surface points are selected to calculate the 3D hand joint coordinates by a weighted average. Experiments demonstrate that NVF outperforms existing state-of-the-art algorithms on FreiHAND dataset for camera-space 3D hand pose estimation. We also adapt NVF to the classic task of root-relative 3D hand pose estimation, for which NVF also obtains state-of-the-art results on HO3D dataset.
Open intent detection is a significant problem in natural language understanding, which aims to identify the unseen open intent while ensuring known intent identification performance. However, current methods face two major challenges. Firstly, they struggle to learn friendly representations to detect the open intent with prior knowledge of only known intents. Secondly, there is a lack of an effective approach to obtaining specific and compact decision boundaries for known intents. To address these issues, this paper presents an original framework called DA-ADB, which successively learns distance-aware intent representations and adaptive decision boundaries for open intent detection. Specifically, we first leverage distance information to enhance the distinguishing capability of the intent representations. Then, we design a novel loss function to obtain appropriate decision boundaries by balancing both empirical and open space risks. Extensive experiments demonstrate the effectiveness of the proposed distance-aware and boundary learning strategies. Compared to state-of-the-art methods, our framework achieves substantial improvements on three benchmark datasets. Furthermore, it yields robust performance with varying proportions of labeled data and known categories.
Image-based rendering techniques stand at the core of an immersive experience for the user, as they generate novel views given a set of multiple input images. Since they have shown good performance in terms of objective and subjective quality, the research community devotes great effort to their improvement. However, the large volume of data necessary to render at the receiver's side hinders applications in limited bandwidth environments or prevents their employment in real-time applications. We present LeHoPP, a method for input pixel pruning, where we examine the importance of each input pixel concerning the rendered view, and we avoid the use of irrelevant pixels. Even without retraining the image-based rendering network, our approach shows a good trade-off between synthesis quality and pixel rate. When tested in the general neural rendering framework, compared to other pruning baselines, LeHoPP gains between $0.9$ dB and $3.6$ dB on average.
Consider an opaque medium which contains 3D particles. All particles are convex bodies of the same shape, but they vary in size. The particles are randomly positioned and oriented within the medium and cannot be observed directly. Taking a planar section of the medium we obtain a sample of observed 2D section profile areas of the intersected particles. In this paper the distribution of interest is the underlying 3D particle size distribution for which an identifiability result is obtained. Moreover, a nonparametric estimator is proposed for this size distribution. The estimator is proven to be consistent and its performance is assessed in a simulation study.
Multi-modal Magnetic Resonance Imaging (MRI) plays an important role in clinical medicine. However, the acquisitions of some modalities, such as the T2-weighted modality, need a long time and they are always accompanied by motion artifacts. On the other hand, the T1-weighted image (T1WI) shares the same underlying information with T2-weighted image (T2WI), which needs a shorter scanning time. Therefore, in this paper we accelerate the acquisition of the T2WI by introducing the auxiliary modality (T1WI). Concretely, we first reconstruct high-quality T2WIs with under-sampled T2WIs. Here, we realize fast T2WI reconstruction by reducing the sampling rate in the k-space. Second, we establish a cross-modal synthesis task to generate the synthetic T2WIs for guiding better T2WI reconstruction. Here, we obtain the synthetic T2WIs by decomposing the whole cross-modal generation mapping into two OT processes, the spatial alignment mapping on the T1 image manifold and the cross-modal synthesis mapping from aligned T1WIs to T2WIs. It overcomes the negative transfer caused by the spatial misalignment. Then, we prove the reconstruction and the synthesis tasks are well complementary. Finally, we compare it with state-of-the-art approaches on an open dataset FastMRI and an in-house dataset to testify the validity of the proposed method.
Event detection tasks can enable the quick detection of events from texts and provide powerful support for downstream natural language processing tasks. Most such methods can only detect a fixed set of predefined event classes. To extend them to detect a new class without losing the ability to detect old classes requires costly retraining of the model from scratch. Incremental learning can effectively solve this problem, but it requires abundant data of new classes. In practice, however, the lack of high-quality labeled data of new event classes makes it difficult to obtain enough data for model training. To address the above mentioned issues, we define a new task, few-shot incremental event detection, which focuses on learning to detect a new event class with limited data, while retaining the ability to detect old classes to the extent possible. We created a benchmark dataset IFSED for the few-shot incremental event detection task based on FewEvent and propose two benchmarks, IFSED-K and IFSED-KP. Experimental results show that our approach has a higher F1-score than baseline methods and is more stable.
When learning tasks over time, artificial neural networks suffer from a problem known as Catastrophic Forgetting (CF). This happens when the weights of a network are overwritten during the training of a new task causing forgetting of old information. To address this issue, we propose MetA Reusable Knowledge or MARK, a new method that fosters weight reusability instead of overwriting when learning a new task. Specifically, MARK keeps a set of shared weights among tasks. We envision these shared weights as a common Knowledge Base (KB) that is not only used to learn new tasks, but also enriched with new knowledge as the model learns new tasks. Key components behind MARK are two-fold. On the one hand, a metalearning approach provides the key mechanism to incrementally enrich the KB with new knowledge and to foster weight reusability among tasks. On the other hand, a set of trainable masks provides the key mechanism to selectively choose from the KB relevant weights to solve each task. By using MARK, we achieve state of the art results in several popular benchmarks, surpassing the best performing methods in terms of average accuracy by over 10% on the 20-Split-MiniImageNet dataset, while achieving almost zero forgetfulness using 55% of the number of parameters. Furthermore, an ablation study provides evidence that, indeed, MARK is learning reusable knowledge that is selectively used by each task.
Existing methods for vision-and-language learning typically require designing task-specific architectures and objectives for each task. For example, a multi-label answer classifier for visual question answering, a region scorer for referring expression comprehension, and a language decoder for image captioning, etc. To alleviate these hassles, in this work, we propose a unified framework that learns different tasks in a single architecture with the same language modeling objective, i.e., multimodal conditional text generation, where our models learn to generate labels in text based on the visual and textual inputs. On 7 popular vision-and-language benchmarks, including visual question answering, referring expression comprehension, visual commonsense reasoning, most of which have been previously modeled as discriminative tasks, our generative approach (with a single unified architecture) reaches comparable performance to recent task-specific state-of-the-art vision-and-language models. Moreover, our generative approach shows better generalization ability on answering questions that have rare answers. In addition, we show that our framework allows multi-task learning in a single architecture with a single set of parameters, which achieves similar performance to separately optimized single-task models. Our code will be publicly available at: //github.com/j-min/VL-T5
This paper focuses on two fundamental tasks of graph analysis: community detection and node representation learning, which capture the global and local structures of graphs, respectively. In the current literature, these two tasks are usually independently studied while they are actually highly correlated. We propose a probabilistic generative model called vGraph to learn community membership and node representation collaboratively. Specifically, we assume that each node can be represented as a mixture of communities, and each community is defined as a multinomial distribution over nodes. Both the mixing coefficients and the community distribution are parameterized by the low-dimensional representations of the nodes and communities. We designed an effective variational inference algorithm which regularizes the community membership of neighboring nodes to be similar in the latent space. Experimental results on multiple real-world graphs show that vGraph is very effective in both community detection and node representation learning, outperforming many competitive baselines in both tasks. We show that the framework of vGraph is quite flexible and can be easily extended to detect hierarchical communities.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191