We study the understanding of embodied reference: One agent uses both language and gesture to refer to an object to another agent in a shared physical environment. Of note, this new visual task requires understanding multimodal cues with perspective-
taking to identify which object is being referred to. To tackle this problem, we introduce YouRefIt, a new crowd-sourced dataset of embodied reference collected in various physical scenes; the dataset contains 4,195 unique reference clips in 432 indoor scenes. To the best of our knowledge, this is the first embodied reference dataset that allows us to study referring expressions in daily physical scenes to understand referential behavior, human communication, and human-robot interaction. We further devise two benchmarks for image-based and video-based embodied reference understanding. Comprehensive baselines and extensive experiments provide the very first result of machine perception on how the referring expressions and gestures affect the embodied reference understanding. Our results provide essential evidence that gestural cues are as critical as language cues in understanding the embodied reference.
Few-shot classification (FSC) is one of the most concerned hot issues in recent years. The general setting consists of two phases: (1) Pre-train a feature extraction model (FEM) with base data (has large amounts of labeled samples). (2) Use the FEM t
o extract the features of novel data (with few labeled samples and totally different categories from base data), then classify them with the to-be-designed classifier. The adaptability of pre-trained FEM to novel data determines the accuracy of novel features, thereby affecting the final classification performances. To this end, how to appraise the pre-trained FEM is the most crucial focus in the FSC community. It sounds like traditional Class Activate Mapping (CAM) based methods can achieve this by overlaying weighted feature maps. However, due to the particularity of FSC (e.g., there is no backpropagation when using the pre-trained FEM to extract novel features), we cannot activate the feature map with the novel classes. To address this challenge, we propose a simple, flexible method, dubbed as Class-Irrelevant Mapping (CIM). Specifically, first, we introduce dictionary learning theory and view the channels of the feature map as the bases in a dictionary. Then we utilize the feature map to fit the feature vector of an image to achieve the corresponding channel weights. Finally, we overlap the weighted feature map for visualization to appraise the ability of pre-trained FEM on novel data. For fair use of CIM in evaluating different models, we propose a new measurement index, called Feature Localization Accuracy (FLA). In experiments, we first compare our CIM with CAM in regular tasks and achieve outstanding performances. Next, we use our CIM to appraise several classical FSC frameworks without considering the classification results and discuss them.
With the fast development of Deep Learning techniques, Named Entity Recognition (NER) is becoming more and more important in the information extraction task. The greatest difficulty that the NER task faces is to keep the detectability even when types
of NE and documents are unfamiliar. Realizing that the specificity information may contain potential meanings of a word and generate semantic-related features for word embedding, we develop a distribution-aware word embedding and implement three different methods to make use of the distribution information in a NER framework. And the result shows that the performance of NER will be improved if the word specificity is incorporated into existing NER methods.
We study the vision transformer structure in the mobile level in this paper, and find a dramatic performance drop. We analyze the reason behind this phenomenon, and propose a novel irregular patch embedding module and adaptive patch fusion module to
improve the performance. We conjecture that the vision transformer blocks (which consist of multi-head attention and feed-forward network) are more suitable to handle high-level information than low-level features. The irregular patch embedding module extracts patches that contain rich high-level information with different receptive fields. The transformer blocks can obtain the most useful information from these irregular patches. Then the processed patches pass the adaptive patch merging module to get the final features for the classifier. With our proposed improvements, the traditional uniform vision transformer structure can achieve state-of-the-art results in mobile level. We improve the DeiT baseline by more than 9% under the mobile-level settings and surpass other transformer architectures like Swin and CoaT by a large margin.
Unlike conventional zero-shot classification, zero-shot semantic segmentation predicts a class label at the pixel level instead of the image level. When solving zero-shot semantic segmentation problems, the need for pixel-level prediction with surrou
nding context motivates us to incorporate spatial information using positional encoding. We improve standard positional encoding by introducing the concept of Relative Positional Encoding, which integrates spatial information at the feature level and can handle arbitrary image sizes. Furthermore, while self-training is widely used in zero-shot semantic segmentation to generate pseudo-labels, we propose a new knowledge-distillation-inspired self-training strategy, namely Annealed Self-Training, which can automatically assign different importance to pseudo-labels to improve performance. We systematically study the proposed Relative Positional Encoding and Annealed Self-Training in a comprehensive experimental evaluation, and our empirical results confirm the effectiveness of our method on three benchmark datasets.
Many meta-learning algorithms can be formulated into an interleaved process, in the sense that task-specific predictors are learned during inner-task adaptation and meta-parameters are updated during meta-update. The normal meta-training strategy nee
ds to differentiate through the inner-task adaptation procedure to optimize the meta-parameters. This leads to a constraint that the inner-task algorithms should be solved analytically. Under this constraint, only simple algorithms with analytical solutions can be applied as the inner-task algorithms, limiting the model expressiveness. To lift the limitation, we propose an adaptation-agnostic meta-training strategy. Following our proposed strategy, we can apply stronger algorithms (e.g., an ensemble of different types of algorithms) as the inner-task algorithm to achieve superior performance comparing with popular baselines. The source code is available at https://github.com/jiaxinchen666/AdaptationAgnosticMetaLearning.
RGB-D based 6D pose estimation has recently achieved remarkable progress, but still suffers from two major limitations: (1) ineffective representation of depth data and (2) insufficient integration of different modalities. This paper proposes a novel
deep learning approach, namely Graph Convolutional Network with Point Refinement (PR-GCN), to simultaneously address the issues above in a unified way. It first introduces the Point Refinement Network (PRN) to polish 3D point clouds, recovering missing parts with noise removed. Subsequently, the Multi-Modal Fusion Graph Convolutional Network (MMF-GCN) is presented to strengthen RGB-D combination, which captures geometry-aware inter-modality correlation through local information propagation in the graph convolutional network. Extensive experiments are conducted on three widely used benchmarks, and state-of-the-art performance is reached. Besides, it is also shown that the proposed PRN and MMF-GCN modules are well generalized to other frameworks.
In second harmonic generation, the phase of the optical field is doubled which has important implication. Here the phase doubling effect is utilized to solve a long-standing challenge in power scaling of single frequency laser. When a (-{pi}/2, {pi}/
2) binary phase modulation is applied to a single frequency seed laser to broaden the spectrum and suppress the stimulated Brillouin scattering in high power fiber amplifier, the second harmonic of the phase-modulated laser will return to single frequency, because the (-{pi}/2, {pi}/2) modulation is doubled to (-{pi}, {pi}) for the second harmonic. A compression rate as high as 95% is demonstrated in the experiment limited by the electronic bandwidth of the setup, which can be improved with optimized devices.
We consider the covariance steering problem for nonlinear control-affine systems. Our objective is to find an optimal control strategy to steer the state of a system from an initial distribution to a target one whose mean and covariance are given. Du
e to the nonlinearity, the existing techniques for linear covariance steering problems are not directly applicable. By leveraging the celebrated Girsanov theorem, we formulate the problem into an optimization over the space path distributions. We then adopt a generalized proximal gradient algorithm to solve this optimization, where each update requires solving a linear covariance steering problem. Our algorithm is guaranteed to converge to a local optimal solution with a sublinear rate. In addition, each iteration of the algorithm can be achieved in closed form, and thus the computational complexity of it is insensitive to the resolution of time-discretization.
Internet video delivery has undergone a tremendous explosion of growth over the past few years. However, the quality of video delivery system greatly depends on the Internet bandwidth. Deep Neural Networks (DNNs) are utilized to improve the quality o
f video delivery recently. These methods divide a video into chunks, and stream LR video chunks and corresponding content-aware models to the client. The client runs the inference of models to super-resolve the LR chunks. Consequently, a large number of models are streamed in order to deliver a video. In this paper, we first carefully study the relation between models of different chunks, then we tactfully design a joint training framework along with the Content-aware Feature Modulation (CaFM) layer to compress these models for neural video delivery. {bf With our method, each video chunk only requires less than $1% $ of original parameters to be streamed, achieving even better SR performance.} We conduct extensive experiments across various SR backbones, video time length, and scaling factors to demonstrate the advantages of our method. Besides, our method can be also viewed as a new approach of video coding. Our primary experiments achieve better video quality compared with the commercial H.264 and H.265 standard under the same storage cost, showing the great potential of the proposed method. Code is available at:url{https://github.com/Neural-video-delivery/CaFM-Pytorch-ICCV2021}
