arXiv subjects

Chao Li

Publications and source records attributed to Chao Li.

At least 379 records · Page 21Linked to original sources

arXiv2022

Instance Segmentation for Chinese Character Stroke Extraction, Datasets and Benchmarks

Stroke is the basic element of Chinese character and stroke extraction has been an important and long-standing endeavor. Existing stroke extraction methods are often handcrafted and highly depend on domain expertise due to the limited training data. Moreover, there are no standardized benchmarks to provide a fair comparison between different stroke extraction methods, which, we believe, is a major impediment to the development of Chinese character stroke understanding and related tasks. In this work, we present the first public available Chinese Character Stroke Extraction (CCSE) benchmark, with two new large-scale datasets: Kaiti CCSE (CCSE-Kai) and Handwritten CCSE (CCSE-HW). With the large-scale datasets, we hope to leverage the representation power of deep models such as CNNs to solve the stroke extraction task, which, however, remains an open question. To this end, we turn the stroke extraction problem into a stroke instance segmentation problem. Using the proposed datasets to train a stroke instance segmentation model, we surpass previous methods by a large margin. Moreover, the models trained with the proposed datasets benefit the downstream font generation and handwritten aesthetic assessment tasks. We hope these benchmark results can facilitate further research. The source code and datasets are publicly available at: https://github.com/lizhaoliu-Lec/CCSE.

cs.CV↗

arXiv2022

Learning from the Dictionary: Heterogeneous Knowledge Guided Fine-tuning for Chinese Spell Checking

Chinese Spell Checking (CSC) aims to detect and correct Chinese spelling errors. Recent researches start from the pretrained knowledge of language models and take multimodal information into CSC models to improve the performance. However, they overlook the rich knowledge in the dictionary, the reference book where one can learn how one character should be pronounced, written, and used. In this paper, we propose the LEAD framework, which renders the CSC model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning. LEAD first constructs positive and negative samples according to the knowledge of character phonetics, glyphs, and definitions in the dictionary. Then a unified contrastive learning-based training scheme is employed to refine the representations of the CSC models. Extensive experiments and detailed analyses on the SIGHAN benchmark datasets demonstrate the effectiveness of our proposed methods.

cs.CL↗

arXiv2022

Rethinking Prototypical Contrastive Learning through Alignment, Uniformity and Correlation

Contrastive self-supervised learning (CSL) with a prototypical regularization has been introduced in learning meaningful representations for downstream tasks that require strong semantic information. However, to optimize CSL with a loss that performs the prototypical regularization aggressively, e.g., the ProtoNCE loss, might cause the "coagulation" of examples in the embedding space. That is, the intra-prototype diversity of samples collapses to trivial solutions for their prototype being well-separated from others. Motivated by previous works, we propose to mitigate this phenomenon by learning Prototypical representation through Alignment, Uniformity and Correlation (PAUC). Specifically, the ordinary ProtoNCE loss is revised with: (1) an alignment loss that pulls embeddings from positive prototypes together; (2) a uniformity loss that distributes the prototypical level features uniformly; (3) a correlation loss that increases the diversity and discriminability between prototypical level features. We conduct extensive experiments on various benchmarks where the results demonstrate the effectiveness of our method in improving the quality of prototypical contrastive representations. Particularly, in the classification down-stream tasks with linear probes, our proposed method outperforms the state-of-the-art instance-wise and prototypical contrastive learning methods on the ImageNet-100 dataset by 2.96% and the ImageNet-1K dataset by 2.46% under the same settings of batch size and epochs.

cs.CV↗

arXiv2022

Scalable photonic integrated circuits for programmable control of atomic systems

Advances in laser technology have driven discoveries in atomic, molecular, and optical (AMO) physics and emerging applications, from quantum computers with cold atoms or ions, to quantum networks with solid-state color centers. This progress is motivating the development of a new generation of "programmable optical control" systems, characterized by criteria (C1) visible (VIS) and near-infrared (IR) wavelength operation, (C2) large channel counts extensible beyond 1000s of individually addressable atoms, (C3) high intensity modulation extinction and (C4) repeatability compatible with low gate errors, and (C5) fast switching times. Here, we address these challenges by introducing an atom control architecture based on VIS-IR photonic integrated circuit (PIC) technology. Based on a complementary metal-oxide-semiconductor (CMOS) fabrication process, this Atom-control PIC (APIC) technology meets the system requirements (C1)-(C5). As a proof of concept, we demonstrate a 16-channel silicon nitride based APIC with (5.8$\pm$0.4) ns response times and -30 dB extinction ratio at a wavelength of 780 nm. This work demonstrates the suitability of PIC technology for quantum control, opening a path towards scalable quantum information processing based on optically-programmable atomic systems.

physics.optics↗

arXiv2022

Mutual Contrastive Low-rank Learning to Disentangle Whole Slide Image Representations for Glioma Grading

Whole slide images (WSI) provide valuable phenotypic information for histological assessment and malignancy grading of tumors. The WSI-based grading promises to provide rapid diagnostic support and facilitate digital health. Currently, the most commonly used WSIs are derived from formalin-fixed paraffin-embedded (FFPE) and Frozen section. The majority of automatic tumor grading models are developed based on FFPE sections, which could be affected by the artifacts introduced by tissue processing. The frozen section exists problems such as low quality that might influence training within single modality as well. To overcome this problem in a single modal training and achieve better multi-modal and discriminative representation disentanglement in brain tumor, we propose a mutual contrastive low-rank learning (MCL) scheme to integrate FFPE and frozen sections for glioma grading. We first design a mutual learning scheme to jointly optimize the model training based on FFPE and frozen sections. In this proposed scheme, we design a normalized modality contrastive loss (NMC-loss), which could promote to disentangle multi-modality complementary representation of FFPE and frozen sections from the same patient. To reduce intra-class variance, and increase inter-class margin at intra- and inter-patient levels, we conduct a low-rank (LR) loss. Our experiments show that the proposed scheme achieves better performance than the model trained based on each single modality or mixed modalities and even improves the feature extraction in classical attention-based multiple instances learning methods (MIL). The combination of NMC-loss and low-rank loss outperforms other typical contrastive loss functions.

eess.IV↗

arXiv2022

AiM: Taking Answers in Mind to Correct Chinese Cloze Tests in Educational Applications

To automatically correct handwritten assignments, the traditional approach is to use an OCR model to recognize characters and compare them to answers. The OCR model easily gets confused on recognizing handwritten Chinese characters, and the textual information of the answers is missing during the model inference. However, teachers always have these answers in mind to review and correct assignments. In this paper, we focus on the Chinese cloze tests correction and propose a multimodal approach (named AiM). The encoded representations of answers interact with the visual information of students' handwriting. Instead of predicting 'right' or 'wrong', we perform the sequence labeling on the answer text to infer which answer character differs from the handwritten content in a fine-grained way. We take samples of OCR datasets as the positive samples for this task, and develop a negative sample augmentation method to scale up the training data. Experimental results show that AiM outperforms OCR-based methods by a large margin. Extensive studies demonstrate the effectiveness of our multimodal approach.

cs.CL↗

arXiv2022

DBT-DMAE: An Effective Multivariate Time Series Pre-Train Model under Missing Data

Multivariate time series(MTS) is a universal data type related to many practical applications. However, MTS suffers from missing data problems, which leads to degradation or even collapse of the downstream tasks, such as prediction and classification. The concurrent missing data handling procedures could inevitably arouse the biased estimation and redundancy-training problem when encountering multiple downstream tasks. This paper presents a universally applicable MTS pre-train model, DBT-DMAE, to conquer the abovementioned obstacle. First, a missing representation module is designed by introducing dynamic positional embedding and random masking processing to characterize the missing symptom. Second, we proposed an auto-encoder structure to obtain the generalized MTS encoded representation utilizing an ameliorated TCN structure called dynamic-bidirectional-TCN as the basic unit, which integrates the dynamic kernel and time-fliping trick to draw temporal features effectively. Finally, the overall feed-in and loss strategy is established to ensure the adequate training of the whole model. Comparative experiment results manifest that the DBT-DMAE outperforms the other state-of-the-art methods in six real-world datasets and two different downstream tasks. Moreover, ablation and interpretability experiments are delivered to verify the validity of DBT-DMAE's substructures.

cs.LG↗

arXiv2022

Omni-Dimensional Dynamic Convolution

Learning a single static convolutional kernel in each convolutional layer is the common training paradigm of modern Convolutional Neural Networks (CNNs). Instead, recent research in dynamic convolution shows that learning a linear combination of $n$ convolutional kernels weighted with their input-dependent attentions can significantly improve the accuracy of light-weight CNNs, while maintaining efficient inference. However, we observe that existing works endow convolutional kernels with the dynamic property through one dimension (regarding the convolutional kernel number) of the kernel space, but the other three dimensions (regarding the spatial size, the input channel number and the output channel number for each convolutional kernel) are overlooked. Inspired by this, we present Omni-dimensional Dynamic Convolution (ODConv), a more generalized yet elegant dynamic convolution design, to advance this line of research. ODConv leverages a novel multi-dimensional attention mechanism with a parallel strategy to learn complementary attentions for convolutional kernels along all four dimensions of the kernel space at any convolutional layer. As a drop-in replacement of regular convolutions, ODConv can be plugged into many CNN architectures. Extensive experiments on the ImageNet and MS-COCO datasets show that ODConv brings solid accuracy boosts for various prevailing CNN backbones including both light-weight and large ones, e.g., 3.77%~5.71%|1.86%~3.72% absolute top-1 improvements to MobivleNetV2|ResNet family on the ImageNet dataset. Intriguingly, thanks to its improved feature learning ability, ODConv with even one single kernel can compete with or outperform existing dynamic convolution counterparts with multiple kernels, substantially reducing extra parameters. Furthermore, ODConv is also superior to other attention modules for modulating the output features or the convolutional weights.

cs.CV↗

arXiv2022

Min-max theory for capillary surfaces

We develop a min-max theory for the construction of capillary surfaces in 3-manifolds with smooth boundary. In particular, for a generic set of ambient metrics, we prove the existence of nontrivial, smooth, almost properly embedded surfaces with any given constant mean curvature $c$, and with smooth boundary contacting at any given constant angle $θ$. Moreover, if $c$ is nonzero and $θ$ is not $\fracπ{2}$, then our min-max solution always has multiplicity one. We also establish a stable Bernstein theorem for minimal hypersurfaces with certain contact angles in higher dimensions.

math.DG↗

arXiv2022

The dihedral rigidity conjecture for n-prisms

We prove the following comparison theorem for metrics with nonnegative scalar curvature, also known as the dihedral rigidity conjecture by Gromov: for $n\le 7$, if an $n$-dimensional prism has nonnegative scalar curvature and weakly mean convex faces, then its dihedral angle cannot be everywhere not larger than its Euclidean model, unless it is isometric to an Euclidean prism. The proof relies on constructing certain free boundary minimal hypersurface in a Riemannian polyhedron, and extending a dimension descent idea of Schoen-Yau. Our result is a localization of the positive mass theorem.

math.DG↗

arXiv2022

MAFormer: A Transformer Network with Multi-scale Attention Fusion for Visual Recognition

Vision Transformer and its variants have demonstrated great potential in various computer vision tasks. But conventional vision transformers often focus on global dependency at a coarse level, which suffer from a learning challenge on global relationships and fine-grained representation at a token level. In this paper, we introduce Multi-scale Attention Fusion into transformer (MAFormer), which explores local aggregation and global feature extraction in a dual-stream framework for visual recognition. We develop a simple but effective module to explore the full potential of transformers for visual representation by learning fine-grained and coarse-grained features at a token level and dynamically fusing them. Our Multi-scale Attention Fusion (MAF) block consists of: i) a local window attention branch that learns short-range interactions within windows, aggregating fine-grained local features; ii) global feature extraction through a novel Global Learning with Down-sampling (GLD) operation to efficiently capture long-range context information within the whole image; iii) a fusion module that self-explores the integration of both features via attention. Our MAFormer achieves state-of-the-art performance on common vision tasks. In particular, MAFormer-L achieves 85.9$\%$ Top-1 accuracy on ImageNet, surpassing CSWin-B and LV-ViT-L by 1.7$\%$ and 0.6$\%$ respectively. On MSCOCO, MAFormer outperforms the prior art CSWin by 1.7$\%$ mAPs on object detection and 1.4$\%$ on instance segmentation with similar-sized parameters, demonstrating the potential to be a general backbone network.

cs.CV↗

arXiv2022

Stimulated Laser Cooling Using Microfabrication

We have achieved stimulated laser cooling of thermal rubidium atomic beams on a silicon chip. Following pre-collimation via a silicon microchannel array, we perform beam brightening via a blue-detuned optical molasses. Owing to the small size of the chip elements, we require only 8 mW, or nine times lower power than earlier free-space experiments on cesium [Aspect et al., Phys. Rev. Lett. 57, 1688 (1986)]. Silicon micromirrors are fabricated and hand-assembled to precisely overlap a strong elliptical standing wave with a sheet-shaped atomic density distribution, with dimensions chosen precisely to match these. We reduce the transverse velocity spread to below 1 m/s within a total travel distance of 4.5 mm on a silicon substrate. We use Doppler-sensitive two-photon Raman spectroscopy to characterize the cooling. In contrast to time-of-flight methods utilized previously, this approach requires a much shorter apparatus to achieve similar resolution. This hybrid of passive and active collimation paves the way toward the construction of full-fledged atomic instruments, such as atomic beam clocks and gyroscopes, entirely on-chip through batch-fabricated processes.

physics.atom-ph↗

arXiv2022

Siamese Prototypical Contrastive Learning

Contrastive Self-supervised Learning (CSL) is a practical solution that learns meaningful visual representations from massive data in an unsupervised approach. The ordinary CSL embeds the features extracted from neural networks onto specific topological structures. During the training progress, the contrastive loss draws the different views of the same input together while pushing the embeddings from different inputs apart. One of the drawbacks of CSL is that the loss term requires a large number of negative samples to provide better mutual information bound ideally. However, increasing the number of negative samples by larger running batch size also enhances the effects of false negatives: semantically similar samples are pushed apart from the anchor, hence downgrading downstream performance. In this paper, we tackle this problem by introducing a simple but effective contrastive learning framework. The key insight is to employ siamese-style metric loss to match intra-prototype features, while increasing the distance between inter-prototype features. We conduct extensive experiments on various benchmarks where the results demonstrate the effectiveness of our method on improving the quality of visual representations. Specifically, our unsupervised pre-trained ResNet-50 with a linear probe, out-performs the fully-supervised trained version on the ImageNet-1K dataset.

cs.CV↗

arXiv2022

Egocentric Activity Recognition and Localization on a 3D Map

Given a video captured from a first person perspective and the environment context of where the video is recorded, can we recognize what the person is doing and identify where the action occurs in the 3D space? We address this challenging problem of jointly recognizing and localizing actions of a mobile user on a known 3D map from egocentric videos. To this end, we propose a novel deep probabilistic model. Our model takes the inputs of a Hierarchical Volumetric Representation (HVR) of the 3D environment and an egocentric video, infers the 3D action location as a latent variable, and recognizes the action based on the video and contextual cues surrounding its potential locations. To evaluate our model, we conduct extensive experiments on the subset of Ego4D dataset, in which both human naturalistic actions and photo-realistic 3D environment reconstructions are captured. Our method demonstrates strong results on both action recognition and 3D action localization across seen and unseen environments. We believe our work points to an exciting research direction in the intersection of egocentric vision, and 3D scene understanding.

cs.CV↗

arXiv2022

Spatiotemporal Self-attention Modeling with Temporal Patch Shift for Action Recognition

Transformer-based methods have recently achieved great advancement on 2D image-based vision tasks. For 3D video-based tasks such as action recognition, however, directly applying spatiotemporal transformers on video data will bring heavy computation and memory burdens due to the largely increased number of patches and the quadratic complexity of self-attention computation. How to efficiently and effectively model the 3D self-attention of video data has been a great challenge for transformers. In this paper, we propose a Temporal Patch Shift (TPS) method for efficient 3D self-attention modeling in transformers for video-based action recognition. TPS shifts part of patches with a specific mosaic pattern in the temporal dimension, thus converting a vanilla spatial self-attention operation to a spatiotemporal one with little additional cost. As a result, we can compute 3D self-attention using nearly the same computation and memory cost as 2D self-attention. TPS is a plug-and-play module and can be inserted into existing 2D transformer models to enhance spatiotemporal feature learning. The proposed method achieves competitive performance with state-of-the-arts on Something-something V1 & V2, Diving-48, and Kinetics400 while being much more efficient on computation and memory cost. The source code of TPS can be found at https://github.com/MartinXM/TPS.

cs.CV↗

arXiv2022

Volume growth of 3-manifolds with scalar curvature lower bounds

We give a new proof of a recent result of Munteanu--Wang relating scalar curvature to volume growth on a $3$-manifold with non-negative Ricci curvature. Our proof relies on the theory of $μ$-bubbles introduced by Gromov as well as the almost splitting theorem due to Cheeger--Colding.

math.DG↗

arXiv2022

SALO: An Efficient Spatial Accelerator Enabling Hybrid Sparse Attention Mechanisms for Long Sequences

The attention mechanisms of transformers effectively extract pertinent information from the input sequence. However, the quadratic complexity of self-attention w.r.t the sequence length incurs heavy computational and memory burdens, especially for tasks with long sequences. Existing accelerators face performance degradation in these tasks. To this end, we propose SALO to enable hybrid sparse attention mechanisms for long sequences. SALO contains a data scheduler to map hybrid sparse attention patterns onto hardware and a spatial accelerator to perform the efficient attention computation. We show that SALO achieves 17.66x and 89.33x speedup on average compared to GPU and CPU implementations, respectively, on typical workloads, i.e., Longformer and ViL.

cs.AR↗

arXiv2022

Security measurement of a medical communication scheme based on chaos and DNA coding

To encrypt sensitive information existing in a color DICOM images, a medical privacy protection scheme (called as MPPS) based on chaos and DNA coding was proposed by using two coupled chaotic systems to produce cryptographic primitives. Relying on some empirical analyses and experimental results, the designers of MPPS claimed that it can withstand a chosen-plaintext attack and some other classic attacking models. However, this statement is groundless. In this paper, we investigate the essential properties of MPPS and DNA coding, and we then propose an efficient chosen-plaintext attack to disclose its equivalent secret-key. The attack only needs $\lceil \log_{256}(3\cdot M\cdot N)\rceil+4$ pair of chosen plain-images and the corresponding cipher-images, where $M \times N$ and ``3" are the size of the RGB color image and the number of color channels, respectively. In addition, the other claimed superiorities are questioned from the perspective of modern cryptography. Both theoretical and experimental results are presented to support the efficiency of the proposed attack and the other reported security faults. The proposed cryptanalysis results will promote the proper application of DNA encoding to protect multimedia privacy data, especially that in a DICOM image.

cs.CR↗