arXiv subjects

Shu Liu

Publications and source records attributed to Shu Liu.

At least 127 records · Page 7Linked to original sources

arXiv2023

Rethinking Out-of-distribution (OOD) Detection: Masked Image Modeling is All You Need

The core of out-of-distribution (OOD) detection is to learn the in-distribution (ID) representation, which is distinguishable from OOD samples. Previous work applied recognition-based methods to learn the ID features, which tend to learn shortcuts instead of comprehensive representations. In this work, we find surprisingly that simply using reconstruction-based methods could boost the performance of OOD detection significantly. We deeply explore the main contributors of OOD detection and find that reconstruction-based pretext tasks have the potential to provide a generally applicable and efficacious prior, which benefits the model in learning intrinsic data distributions of the ID dataset. Specifically, we take Masked Image Modeling as a pretext task for our OOD detection framework (MOOD). Without bells and whistles, MOOD outperforms previous SOTA of one-class OOD detection by 5.7%, multi-class OOD detection by 3.0%, and near-distribution OOD detection by 2.1%. It even defeats the 10-shot-per-class outlier exposure OOD detection, although we do not include any OOD samples for our detection

cs.CV↗

arXiv2023

Point2Pix: Photo-Realistic Point Cloud Rendering via Neural Radiance Fields

Synthesizing photo-realistic images from a point cloud is challenging because of the sparsity of point cloud representation. Recent Neural Radiance Fields and extensions are proposed to synthesize realistic images from 2D input. In this paper, we present Point2Pix as a novel point renderer to link the 3D sparse point clouds with 2D dense image pixels. Taking advantage of the point cloud 3D prior and NeRF rendering pipeline, our method can synthesize high-quality images from colored point clouds, generally for novel indoor scenes. To improve the efficiency of ray sampling, we propose point-guided sampling, which focuses on valid samples. Also, we present Point Encoding to build Multi-scale Radiance Fields that provide discriminative 3D point features. Finally, we propose Fusion Encoding to efficiently synthesize high-quality images. Extensive experiments on the ScanNet and ArkitScenes datasets demonstrate the effectiveness and generalization.

cs.CV↗

arXiv2023

Imaging Molecular Outflow in Massive Star-forming Regions with HNCO Lines

Protostellar outflows are considered a signpost of star formation. These outflows can cause shocks in the molecular gas and are typically traced by the line wings of certain molecules. HNCO (4--3) has been regarded as a shock tracer because of the high abundance in shocked regions. Here we present the first imaging results of HNCO (4--3) line wings toward nine sources in a sample of twenty three massive star-forming regions using the IRAM 30\,m telescope. We adopt the velocity range of the full width of HC$_{3}$N (10--9) and H$^{13}$CO$^+$ (1--0) emissions as the central emission values, beyond which the emission from HNCO (4--3) is considered to be from line wings. The spatial distributions of the red- and/or blue-lobes of HNCO (4--3) emission nicely associate with those lobes of HCO$^{+}$ (1--0) in most of the sources. High intensity ratios of HNCO (4--3) to HCO$^+$ (1--0) are obtained in the line wings. The derived column density ratios of HNCO to HCO$^+$ are consistent with those previously observed towards massive star-forming regions. These results provide direct evidence that HNCO could trace outflow in massive star-forming regions. This work also implies that the formation of some HNCO molecules is related to shock, either on the grain surface or within the shocked gas.

astro-ph.GA↗

arXiv2023

Hierarchical Dense Correlation Distillation for Few-Shot Segmentation

Few-shot semantic segmentation (FSS) aims to form class-agnostic models segmenting unseen classes with only a handful of annotations. Previous methods limited to the semantic feature and prototype representation suffer from coarse segmentation granularity and train-set overfitting. In this work, we design Hierarchically Decoupled Matching Network (HDMNet) mining pixel-level support correlation based on the transformer architecture. The self-attention modules are used to assist in establishing hierarchical dense features, as a means to accomplish the cascade matching between query and support features. Moreover, we propose a matching module to reduce train-set overfitting and introduce correlation distillation leveraging semantic correspondence from coarse resolution to boost fine-grained segmentation. Our method performs decently in experiments. We achieve $50.0\%$ mIoU on \coco~dataset one-shot setting and $56.0\%$ on five-shot segmentation, respectively.

cs.CV↗

arXiv2023

Learning Context-aware Classifier for Semantic Segmentation

Semantic segmentation is still a challenging task for parsing diverse contexts in different scenes, thus the fixed classifier might not be able to well address varying feature distributions during testing. Different from the mainstream literature where the efficacy of strong backbones and effective decoder heads has been well studied, in this paper, additional contextual hints are instead exploited via learning a context-aware classifier whose content is data-conditioned, decently adapting to different latent distributions. Since only the classifier is dynamically altered, our method is model-agnostic and can be easily applied to generic segmentation models. Notably, with only negligible additional parameters and +2\% inference time, decent performance gain has been achieved on both small and large models with challenging benchmarks, manifesting substantial practical merits brought by our simple yet effective method. The implementation is available at \url{https://github.com/tianzhuotao/CAC}.

cs.CV↗

arXiv2023

Discovery of a New Molecular Bubble-Outflow Structure in the Taurus B18 Cloud

Star formation can produce bubbles and outflows, as a result of stellar feedback. Outflows and bubbles inject momentum and energy into the surrounding interstellar medium, and so are related to the overall energy balance of the molecular cloud. Molecular bubbles can be resolved by higher-resolution radio telescopes to quantify the effect of star formation on molecular clouds. We report here the identification of a new molecular bubble with an outflow, and an Herbig Haro object, HH319, located at the bubble center. Multi-wavelength data have been utilized to study its spatial structure, energy injection, and dynamical timescale. This bubble has a kinetic energy of $\rm 5.8 \times 10^{43}$ erg within the smallest radius of a bubble in Taurus, 0.077 pc. The bubble formed $\sim$70,000 years ago. According to the proper motion velocities of protostars from $Gaia$ EDR3, the T Tauri binary stars (FY Tau and FZ Tau) at the southwest edge of the bubble may have produced the outflow-bubble structure. This is an unusual new structure found in low- and intermediate-mass star formation regions. Only a bubble in Orion A, driven by V380 Ori, has a similar structure. The bubble-outflow structure provides additional observational evidence for the theory of stellar wind from T Tauri stars. It enhances our understanding of how stellar feedback acts on molecular clouds.

astro-ph.GA↗

arXiv2022

Neural Monge Map estimation and its applications

Monge map refers to the optimal transport map between two probability distributions and provides a principled approach to transform one distribution to another. Neural network based optimal transport map solver has gained great attention in recent years. Along this line, we present a scalable algorithm for computing the neural Monge map between two probability distributions. Our algorithm is based on a weak form of the optimal transport problem, thus it only requires samples from the marginals instead of their analytic expressions, and can accommodate optimal transport between two distributions with different dimensions. Our algorithm is suitable for general cost functions, compared with other existing methods for estimating Monge maps using samples, which are usually for quadratic costs. The performance of our algorithms is demonstrated through a series of experiments with both synthetic and realistic data, including text-to-image generation and image inpainting tasks.

cs.LG↗

arXiv2022

Non-detection of Broad Hydrogen Radio Recombination Lines in Circinus Galaxy

The line widths of broad line regions (BLRs) of AGNs are key parameters for understanding the central super massive black holes (SMBHs). However, due to obscuration from dusty torus, optical recombination lines from BLRs in type II AGNs can not be directly detected. Radio recombination lines (RRLs), with low extinction, can be ideal tracers to probe emission from BLRs in type II AGNs. We performed RRL observations for H35$α$ and H36$α$ toward the center of Circinus galaxy with ALMA. Narrow components of H35$α$ and H36$α$, which are thought to be mainly from star forming regions around the nuclear region, are detected. However, only upper limits are obtained for broad H35$α$ and H36$α$. Since Circinus galaxy is one of the nearest AGN, non-detection of broad RRLs in Circinus galaxy at this band tells us that it is hopeless to detect broad RRL emission in local AGNs with current facilities. Submillimetre RRLs, with flux densities that are dozens of times higher than those at the millimetre level, could be the tools to directly detect BLRs in type II AGNs with ALMA, once its backend frequency coverage has been upgraded to several times better than its current capabilities.

astro-ph.GA↗

arXiv2022

Optimal control for stochastic nonlinear Schrodinger equation on graph

We study the optimal control formulation for stochastic nonlinear Schrodinger equation (SNLSE) on a finite graph. By viewing the SNLSE as a stochastic Wasserstein Hamiltonian flow on density manifold, we show the global existence of a unique strong solution for SNLSE with a linear drift control or a linear diffusion control on graph. Furthermore, we provide the gradient formula, the existence of the optimal control and a description on the optimal condition via the forward and backward stochastic differential equations.

math.OC↗

arXiv2022

A new construction of nonlinear codes via algebraic function fields

In coding theory, constructing codes with good parameters is one of the most important and fundamental problems. Though a great many of good codes have been produced, most of them are defined over alphabets of sizes equal to prime powers. In this paper, we provide a new explicit construction of $(q+1)$-ary nonlinear codes via algebraic function fields, where $q$ is a prime power. Our codes are constructed by evaluations of rational functions at all rational places of the algebraic function field. Compared with algebraic geometry codes, the main difference is that we allow rational functions to be evaluated at pole places. After evaluating rational functions from a union of Riemann-Roch spaces, we obtain a family of nonlinear codes over the alphabet $\mathbb{F}_{q}\cup \{\infty\}$. It turns out that our codes have better parameters than those obtained from MDS codes or good algebraic geometry codes via code alphabet extension and restriction.

cs.IT↗

arXiv2022

Good locally repairable codes via propagation rules

In classical coding theory, it is common to construct new codes via propagation rules. There are various propagation rules to construct classical block codes. However, propagation rules have not been extensively explored for constructions of locally repairable codes. In this paper, we introduce a few propagation rules to construct good locally repairable codes. To our surprise, these simple propagation rules produce a few interesting results. Firstly, by concatenating a locally repairable code as an inner code with a classical block code as an outer code, we obtain quite a few dimension-optimal binary locally repairable codes. Secondly, from this concatenation, we explicitly build a family of locally repairable codes that exceeds the Zyablov-type bound. Thirdly, by a lengthening propagation rule that adds some rows and columns from a parity-check matrix of a given linear code, we are able to produce a family of dimension-optimal binary locally repairable codes from the extended Hamming codes, and to convert a classical maximum distance separable (MDS) code into a Singleton-optimal locally repairable code. Furthermore, via the lengthening propagation rule, we greatly simplify the construction of a family of locally repairable codes in \cite[Theorem 5]{MX20} that breaks the asymptotic Gilbert-Varshamov bound. In addition, we make use of three other propagation rules to produce more dimension-optimal binary locally repairable codes. Finally, one of phenomena that we observe in this paper is that some trivial propagation rules in classical block codes do not hold anymore for locally repairable codes.

cs.IT↗

arXiv2022

Properties of dense molecular gas along the major axis of M 82

Dense gas is important for galaxy evolution and star formation. Optically-thin dense-gas tracers, such as isotopologues of HCN, HCO+, etc., are very helpful to diagnose excitation conditions of dense molecular gas. However, previous studies of optically-thin dense-gas tracers were mostly focusing on average properties of galaxies as a whole, due to limited sensitivity and angular resolution. M82, a nearby prototype starburst galaxy, offers a unique case for spatially-resolved studies with single-dish telescopes. With the IRAM 30-m telescope, we observed the J = 1 - 0 transition of H13CN, HC15N, H13CO+, HN13C, H15NC, and SiO J = 2 - 1, HC3N J= 10 - 9, H2CO J = 2 - 1 toward five positions along the major axis of M82. The intensity ratios of I(HCN)/I(H13CN) and I(HCO+)/I(H13CO+) show a significant spatial variation along the major axis, with lower values in the central region than those on the disk, indicating higher optical depths in the central region. The optical depths of HCO+ lines are found to be systematically higher than those of HCN lines at all positions. Futhermore, we find that the 14N/15N ratios have an increasing gradient from the center to the outer disk.

astro-ph.GA↗

arXiv2022

DSGN++: Exploiting Visual-Spatial Relation for Stereo-based 3D Detectors

Camera-based 3D object detectors are welcome due to their wider deployment and lower price than LiDAR sensors. We first revisit the prior stereo detector DSGN for its stereo volume construction ways for representing both 3D geometry and semantics. We polish the stereo modeling and propose the advanced version, DSGN++, aiming to enhance effective information flow throughout the 2D-to-3D pipeline in three main aspects. First, to effectively lift the 2D information to stereo volume, we propose depth-wise plane sweeping (DPS) that allows denser connections and extracts depth-guided features. Second, for grasping differently spaced features, we present a novel stereo volume -- Dual-view Stereo Volume (DSV) that integrates front-view and top-view features and reconstructs sub-voxel depth in the camera frustum. Third, as the foreground region becomes less dominant in 3D space, we propose a multi-modal data editing strategy -- Stereo-LiDAR Copy-Paste, which ensures cross-modal alignment and improves data efficiency. Without bells and whistles, extensive experiments in various modality setups on the popular KITTI benchmark show that our method consistently outperforms other camera-based 3D detectors for all categories. Code is available at https://github.com/chenyilun95/DSGN2.

cs.CV↗

arXiv2022

Improving Data Driven Inverse Text Normalization using Data Augmentation

Inverse text normalization (ITN) is used to convert the spoken form output of an automatic speech recognition (ASR) system to a written form. Traditional handcrafted ITN rules can be complex to transcribe and maintain. Meanwhile neural modeling approaches require quality large-scale spoken-written pair examples in the same or similar domain as the ASR system (in-domain data), to train. Both these approaches require costly and complex annotations. In this paper, we present a data augmentation technique that effectively generates rich spoken-written numeric pairs from out-of-domain textual data with minimal human annotation. We empirically demonstrate that ITN model trained using our data augmentation technique consistently outperform ITN model trained using only in-domain data across all numeric surfaces like cardinal, currency, and fraction, by an overall accuracy of 14.44%.

cs.CL↗

arXiv2022

DecoupleNet: Decoupled Network for Domain Adaptive Semantic Segmentation

Unsupervised domain adaptation in semantic segmentation has been raised to alleviate the reliance on expensive pixel-wise annotations. It leverages a labeled source domain dataset as well as unlabeled target domain images to learn a segmentation network. In this paper, we observe two main issues of the existing domain-invariant learning framework. (1) Being distracted by the feature distribution alignment, the network cannot focus on the segmentation task. (2) Fitting source domain data well would compromise the target domain performance. To address these issues, we propose DecoupleNet that alleviates source domain overfitting and enables the final model to focus more on the segmentation task. Furthermore, we put forward Self-Discrimination (SD) and introduce an auxiliary classifier to learn more discriminative target domain features with pseudo labels. Finally, we propose Online Enhanced Self-Training (OEST) to contextually enhance the quality of pseudo labels in an online manner. Experiments show our method outperforms existing state-of-the-art methods, and extensive ablation studies verify the effectiveness of each component. Code is available at https://github.com/dvlab-research/DecoupleNet.

cs.CV↗

arXiv2022

Upper bounds on maximum lengths of Singleton-optimal locally repairable codes

A locally repairable code is called Singleton-optimal if it achieves the Singleton-type bound. Such codes are of great theoretic interest in the study of locally repairable codes. In the recent years there has been a great amount of work on this topic. One of the main problems in this topic is to determine the largest length of a q-ary Singleton-optimal locally repairable code for given locality and minimum distance. Unlike classical MDS codes, the maximum length of Singleton? Optimal locally repairable codes are very sensitive to minimum distance and locality. Thus, it is more challenging and complicated to investigate the maximum length of Singleton-optimal locally repairable codes. In literature, there has been already some research on this problem. However, most of work is concerned with some specific parameter regime such as small minimum distance and locality, and rely on the constraint that (r + 1)|n and recovery sets are disjoint, where r is locality and n is the code length. In this paper we study the problem for large range of parameters including the case where minimum distance is proportional to length. In addition, we also derive some upper bounds on the maximum length of Singleton-optimal locally repairable codes with small minimum distance by removing this constraint. It turns out that even without the constraint we still get better upper bounds for codes with small locality and distance compared with known results. Furthermore, based on our upper bounds for codes with small distance and locality and some propagation rule that we propose in this paper, we are able to derive some upper bounds for codes with relatively large distance and locality assuming that (r + 1)|n and recovery sets are disjoint.

cs.IT↗

arXiv2022

Neural Parametric Fokker-Planck Equations

In this paper, we develop and analyze numerical methods for high dimensional Fokker-Planck equations by leveraging generative models from deep learning. Our starting point is a formulation of the Fokker-Planck equation as a system of ordinary differential equations (ODEs) on finite-dimensional parameter space with the parameters inherited from generative models such as normalizing flows. We call such ODEs neural parametric Fokker-Planck equations. The fact that the Fokker-Planck equation can be viewed as the $L^2$-Wasserstein gradient flow of Kullback-Leibler (KL) divergence allows us to derive the ODEs as the constrained $L^2$-Wasserstein gradient flow of KL divergence on the set of probability densities generated by neural networks. For numerical computation, we design a variational semi-implicit scheme for the time discretization of the proposed ODE. Such an algorithm is sampling-based, which can readily handle the Fokker-Planck equations in higher dimensional spaces. Moreover, we also establish bounds for the asymptotic convergence analysis of the neural parametric Fokker-Planck equation as well as the error analysis for both the continuous and discrete versions. Several numerical examples are provided to illustrate the performance of the proposed algorithms and analysis.

math.NA↗

arXiv2022

EfficientNeRF: Efficient Neural Radiance Fields

Neural Radiance Fields (NeRF) has been wildly applied to various tasks for its high-quality representation of 3D scenes. It takes long per-scene training time and per-image testing time. In this paper, we present EfficientNeRF as an efficient NeRF-based method to represent 3D scene and synthesize novel-view images. Although several ways exist to accelerate the training or testing process, it is still difficult to much reduce time for both phases simultaneously. We analyze the density and weight distribution of the sampled points then propose valid and pivotal sampling at the coarse and fine stage, respectively, to significantly improve sampling efficiency. In addition, we design a novel data structure to cache the whole scene during testing to accelerate the rendering speed. Overall, our method can reduce over 88\% of training time, reach rendering speed of over 200 FPS, while still achieving competitive accuracy. Experiments prove that our method promotes the practicality of NeRF in the real world and enables many applications.

cs.CV↗