Search arXiv⌕ Search

arXiv subjects

Yue Ma

Publications and source records attributed to Yue Ma.

At least 145 records · Page 8Linked to original sources
arXiv2024

SemanticSLAM: Learning based Semantic Map Construction and Robust Camera Localization

Current techniques in Visual Simultaneous Localization and Mapping (VSLAM) estimate camera displacement by comparing image features of consecutive scenes. These algorithms depend on scene continuity, hence requires frequent camera inputs. However, processing images frequently can lead to significant memory usage and computation overhead. In this study, we introduce SemanticSLAM, an end-to-end visual-inertial odometry system that utilizes semantic features extracted from an RGB-D sensor. This approach enables the creation of a semantic map of the environment and ensures reliable camera localization. SemanticSLAM is scene-agnostic, which means it doesn't require retraining for different environments. It operates effectively in indoor settings, even with infrequent camera input, without prior knowledge. The strength of SemanticSLAM lies in its ability to gradually refine the semantic map and improve pose estimation. This is achieved by a convolutional long-short-term-memory (ConvLSTM) network, trained to correct errors during map construction. Compared to existing VSLAM algorithms, SemanticSLAM improves pose estimation by 17%. The resulting semantic map provides interpretable information about the environment and can be easily applied to various downstream tasks, such as path planning, obstacle avoidance, and robot navigation. The code will be publicly available at https://github.com/Leomingyangli/SemanticSLAM

cs.RO↗
arXiv2024

Follow Your Pose: Pose-Guided Text-to-Video Generation using Pose-Free Videos

Generating text-editable and pose-controllable character videos have an imperious demand in creating various digital human. Nevertheless, this task has been restricted by the absence of a comprehensive dataset featuring paired video-pose captions and the generative prior models for videos. In this work, we design a novel two-stage training scheme that can utilize easily obtained datasets (i.e.,image pose pair and pose-free video) and the pre-trained text-to-image (T2I) model to obtain the pose-controllable character videos. Specifically, in the first stage, only the keypoint-image pairs are used only for a controllable text-to-image generation. We learn a zero-initialized convolutional encoder to encode the pose information. In the second stage, we finetune the motion of the above network via a pose-free video dataset by adding the learnable temporal self-attention and reformed cross-frame self-attention blocks. Powered by our new designs, our method successfully generates continuously pose-controllable character videos while keeps the editing and concept composition ability of the pre-trained T2I model. The code and models will be made publicly available.

cs.CV↗
arXiv2023

M-BEV: Masked BEV Perception for Robust Autonomous Driving

3D perception is a critical problem in autonomous driving. Recently, the Bird-Eye-View (BEV) approach has attracted extensive attention, due to low-cost deployment and desirable vision detection capacity. However, the existing models ignore a realistic scenario during the driving procedure, i.e., one or more view cameras may be failed, which largely deteriorates the performance. To tackle this problem, we propose a generic Masked BEV (M-BEV) perception framework, which can effectively improve robustness to this challenging scenario, by random masking and reconstructing camera views in the end-to-end training. More specifically, we develop a novel Masked View Reconstruction (MVR) module for M-BEV. It mimics various missing cases by randomly masking features of different camera views, then leverages the original features of these views as self-supervision, and reconstructs the masked ones with the distinct spatio-temporal context across views. Via such a plug-and-play MVR, our M-BEV is capable of learning the missing views from the resting ones, and thus well generalized for robust view recovery and accurate perception in the testing. We perform extensive experiments on the popular NuScenes benchmark, where our framework can significantly boost 3D perception performance of the state-of-the-art models on various missing view cases, e.g., for the absence of back view, our M-BEV promotes the PETRv2 model with 10.3% mAP gain.

cs.CV↗
arXiv2023

MagicScroll: Nontypical Aspect-Ratio Image Generation for Visual Storytelling via Multi-Layered Semantic-Aware Denoising

Visual storytelling often uses nontypical aspect-ratio images like scroll paintings, comic strips, and panoramas to create an expressive and compelling narrative. While generative AI has achieved great success and shown the potential to reshape the creative industry, it remains a challenge to generate coherent and engaging content with arbitrary size and controllable style, concept, and layout, all of which are essential for visual storytelling. To overcome the shortcomings of previous methods including repetitive content, style inconsistency, and lack of controllability, we propose MagicScroll, a multi-layered, progressive diffusion-based image generation framework with a novel semantic-aware denoising process. The model enables fine-grained control over the generated image on object, scene, and background levels with text, image, and layout conditions. We also establish the first benchmark for nontypical aspect-ratio image generation for visual storytelling including mediums like paintings, comics, and cinematic panoramas, with customized metrics for systematic evaluation. Through comparative and ablation studies, MagicScroll showcases promising results in aligning with the narrative text, improving visual coherence, and engaging the audience. We plan to release the code and benchmark in the hope of a better collaboration between AI researchers and creative practitioners involving visual storytelling.

cs.CV↗
arXiv2023

Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection

Video Moment Retrieval (MR) and Highlight Detection (HD) have attracted significant attention due to the growing demand for video analysis. Recent approaches treat MR and HD as similar video grounding problems and address them together with transformer-based architecture. However, we observe that the emphasis of MR and HD differs, with one necessitating the perception of local relationships and the other prioritizing the understanding of global contexts. Consequently, the lack of task-specific design will inevitably lead to limitations in associating the intrinsic specialty of two tasks. To tackle the issue, we propose a Unified Video COMprehension framework (UVCOM) to bridge the gap and jointly solve MR and HD effectively. By performing progressive integration on intra and inter-modality across multi-granularity, UVCOM achieves the comprehensive understanding in processing a video. Moreover, we present multi-aspect contrastive learning to consolidate the local relation modeling and global knowledge accumulation via well aligned multi-modal space. Extensive experiments on QVHighlights, Charades-STA, TACoS , YouTube Highlights and TVSum datasets demonstrate the effectiveness and rationality of UVCOM which outperforms the state-of-the-art methods by a remarkable margin.

cs.CV↗
arXiv2023

On the saturation spectrum of the unions of disjoint cycles

Let $G$ be a graph and $\mathcal{H}$ be a family of graphs. We say $G$ is $\mathcal{H}$-saturated if $G$ does not contain a copy of $H$ with $H\in\mathcal{H}$, but the addition of any edge $e\notin E(G)$ creates at least one copy of some $H\in\mathcal{H}$ within $G+e$. The saturation number of $\mathcal{H}$ is the minimum size of an $\mathcal{H}$-saturated graph on $n$ vertices, and the saturation spectrum of $\mathcal{H}$ is the set of all possible sizes of an $\mathcal{H}$-saturated graph on $n$ vertices. Let $k\mathcal{C}_{\ge 3}$ be the family of the unions of $k$ vertex-disjoint cycles. In this note, we completely determine the saturation number and the saturation spectrum of $k\mathcal{C}_{\ge 3}$ for $k=2$ and give some results for $k\ge 3$.

math.CO↗
arXiv2023

Coupling undetected sensing modes by quantum erasure

The effect known as ``induced coherence without induced emission'' has spawned a field dedicated to imaging with undetected photons (IUP), where photons from two distinct photon-pair sources interfere if their outputs are made indistinguishable. The indistinguishability is commonly achieved in two setups. Induced coherence IUP (IC-IUP) has only the idler photons from the first source passing through the second, whilst nonlinear interferometry (NI-IUP) has both signal and idler photons from the first source passing through the second and can be simpler to implement. In both cases, changes in the idler path between sources can be detected by measuring the interference fringes in the signal path in a way that allows image information to be moved between different wavelengths. Here we model and implement a novel setup that uses a polarization state quantum eraser approach to move continuously between IC-IUP and NI-IUP operation. We find excellent agreement between experiment and theory in the low-gain or quantum regime. The system also provides a new route for optimizing IUP interference by using controllable quantum erasure to balance the interferometer.

quant-ph↗
arXiv2023

Conflating point of interest (POI) data: A systematic review of matching methods

Point of interest (POI) data provide digital representations of places in the real world, and have been increasingly used to understand human-place interactions, support urban management, and build smart cities. Many POI datasets have been developed, which often have different geographic coverages, attribute focuses, and data quality. From time to time, researchers may need to conflate two or more POI datasets in order to build a better representation of the places in the study areas. While various POI conflation methods have been developed, there lacks a systematic review, and consequently, it is difficult for researchers new to POI conflation to quickly grasp and use these existing methods. This paper fills such a gap. Following the protocol of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), we conduct a systematic review by searching through three bibliographic databases using reproducible syntax to identify related studies. We then focus on a main step of POI conflation, i.e., POI matching, and systematically summarize and categorize the identified methods. Current limitations and future opportunities are discussed afterwards. We hope that this review can provide some guidance for researchers interested in conflating POI datasets for their research.

cs.CY↗
arXiv2023

The degree threshold for covering with all the connected $3$-graphs with $3$ edges

Given two $r$-uniform hypergraphs $F$ and $H$, we say that $H$ has an $F$-covering if every vertex in $H$ is contained in a copy of $F$. Let $c_{i}(n,F)$ be the least integer such that every $n$-vertex $r$-graph $H$ with $δ_{i}(H)>c_i(n,F)$ has an $F$-covering. Falgas-Ravry, Markstöm and Zhao (Combin. Probab. Comput., 2021) asymptotically determined $c_1(n,K_{4}^{(3)-})$, where $K_{4}^{(3)-}$ is obtained by deleting an edge from the complete $3$-graph on $4$ vertices. Later, Tang, Ma and Hou (arXiv, 2022) asymptotically determined $c_1(n,C_{6}^{(3)})$, where $C_{6}^{(3)}$ is the linear triangle, i.e. $C_{6}^{(3)}=([6],\{123,345,561\})$. In this paper, we determine $c_1(n,F_5)$ asymptotically, where $F_5$ is the generalized triangle, i.e. $F_5=([5],\{123,124,345\})$. We also determine the exact values of $c_1(n,F)$, where $F$ is any connected $3$-graphs with $3$ edges and $F\notin\{K_4^{(3)-}, C_{6}^{(3)}, F_5\}$.

math.CO↗
arXiv2023

Graphs without rainbow cliques of orders four and five

Let $\mathcal{G}_n^k=\{G_1,G_2,\ldots,G_k\}$ be a multiset of graphs on vertex set $[n]$ and let $F$ be a fixed graph with edge set $F=\{e_1, e_2,\ldots, e_m\}$ and $k\ge m$. We say ${\mathcal{G}_n^k}$ is rainbow $F$-free if there is no $\{i_1, i_2,\ldots, i_{m}\}\subseteq[k]$ satisfying $e_j\in G_{i_j}$ for every $j\in[m]$. Let $\ex_k(n,F)$ be the maximum $\sum_{i=1}^{k}|G_i|$ among all the rainbow $F$-free multisets ${\mathcal{G}_n^k}$. Keevash, Saks, Sudakov, and Verstraëte (2004) determined the exact value of $\ex_k(n, K_r)$ when $n$ is sufficiently large and proposed the conjecture that the results remain true when $n\ge Cr^2$ for some constant $C$. Recently, Frankl (2022) confirmed the conjecture for $r=3$ and all possible values of $n$. In this paper, we determine the exact value of $\ex_k(n, K_r)$ for $n\ge r-1$ when $r=4$ and $5$, i.e. the conjecture of Keevash, Saks, Sudakov, and Verstraëte is true for $r\in\{4,5\}$.

math.CO↗
arXiv2023

Low Energy Electronic Recoils and Single Electron Detection with a Liquid Xenon Proportional Scintillation Counter

Liquid xenon (LXe) is a well-studied detector medium to search for rare events in dark matter and neutrino physics. Two-phase xenon time projection chambers (TPCs) can detect electronic and nuclear recoils with energy down to kilo-electron volts (keV). In this paper, we characterize the response of a single-phase liquid xenon proportional scintillation counter (LXePSC), which produces electroluminescence directly in the liquid, to detect electronic recoils at low energies. Our design uses a thin (10 - 25 $μ$m diameter), central anode wire in a cylindrical LXe target where ionization electrons, created from radiation particles, drift radially towards the anode, and electroluminescence is produced. Both the primary scintillation (S1) and electroluminescence (S2) are detected by photomultiplier tubes (PMTs) surrounding the LXe target. Up to 17 photons are produced per electron, obtained with a 10 $μ$m diameter anode wire, allowing for the highly efficient detection of electronic recoils from beta decays of a tritium source down to roughly 1 keV. Single electrons, from photo-emission of the cathode wires, are observed at a gain of 1.8 photoelectrons (PE) per electron. The delayed signals following the S2 signals are dominated by single-photon-like hits, without evidence for electron signals observed in the two-phase xenon TPCs. We discuss the potential application of such a LXePSC for reactor neutrino detection via Coherent Elastic Neutrino Nucleus Scattering (CE$ν$NS).

physics.ins-det↗
arXiv2023

Boundary Blending: Reconsidering the Design of Multi-View Visualizations

Multiple-view visualizations (MVs) have been widely used for visual analysis. Each view shows some part of the data in a usable way, and together multiple views enable a holistic understanding of the data under investigation. For example, an analyst may check a social network graph, a map of sensitive locations, a table of transaction records, and a collection of reports to identify suspicious activities. While each view is designed to preserve its own visual context with visible borders or perceivable spatial distance from others, the key to solving real-world analysis problems often requires "breaking" such boundaries, and further integrating and synthesizing the data scattered across multiple views. This calls for blending the boundaries in MVs, instead of simply breaking them, which brings key questions: what are possible boundaries in MVs, and what are design options that can support the boundary blending in MVs? To answer these questions, we present three boundaries in MVs: 1) data boundary, 2) representation boundary, and 3) semantic boundary, corresponding to three major aspects regarding the usage of MVs: encoded information, visual representation, and interpretation. Then, we discuss four design strategies (highlighting, linking, embedding, and extending) and their pros and cons for supporting boundary blending in MVs. We conclude our discussion with future research opportunities.

cs.HC↗
arXiv2023

Efficient Computation of General Modules for ALC Ontologies (Extended Version)

We present a method for extracting general modules for ontologies formulated in the description logic ALC. A module for an ontology is an ideally substantially smaller ontology that preserves all entailments for a user-specified set of terms. As such, it has applications such as ontology reuse and ontology analysis. Different from classical modules, general modules may use axioms not explicitly present in the input ontology, which allows for additional conciseness. So far, general modules have only been investigated for lightweight description logics. We present the first work that considers the more expressive description logic ALC. In particular, our contribution is a new method based on uniform interpolation supported by some new theoretical results. Our evaluation indicates that our general modules are often smaller than classical modules and uniform interpolants computed by the state-of-the-art, and compared with uniform interpolants, can be computed in a significantly shorter time. Moreover, our method can be used for, and in fact improves, the computation of uniform interpolants and classical modules.

cs.AI↗
arXiv2023

Non-Pauli errors can be efficiently sampled in qudit surface codes

Surface codes are the most promising candidates for fault-tolerant quantum computation. Single qudit errors are typically modelled as Pauli operators, to which general errors are converted via randomizing methods. In this Letter, we quantify remaining correlations after syndrome measurement for a qudit 2D surface code subject to non-Pauli errors. Using belief propagation and percolation theory, we relate correlations to loops on the lattice. Below the error correction threshold, remaining correlations are sparse and locally constrained. Syndromes for qudit surface codes are therefore efficiently samplable for non-Pauli errors, independent of the exact forms of the error and decoder.

quant-ph↗
arXiv2023

Eliminating Thermal IR Background Noise by Imaging with Undetected Photons

Spectroscopy and imaging in the mid-infrared (2.5 $μ$m $\sim$ $λ$ $\sim$ 25 $μ$m) is bedevilled by the presence of a strong 300 K thermal background at room temperature that makes IR detectors decades noisier than can be readily achieved in the visible. The technique of "imaging with undetected photons" (IUP) exploits the quantum correlations between entangled photon pairs to transfer image information from one spectral region to another, and here we show that it does so in a way that is immune to the thermal background. This means that IUP can be used to perform high speed photon counting measurements across the mid-IR, using uncooled visible detectors that are many times cheaper, faster, and more sensitive than their IR counterparts.

quant-ph↗
arXiv2023

SemanticAC: Semantics-Assisted Framework for Audio Classification

In this paper, we propose SemanticAC, a semantics-assisted framework for Audio Classification to better leverage the semantic information. Unlike conventional audio classification methods that treat class labels as discrete vectors, we employ a language model to extract abundant semantics from labels and optimize the semantic consistency between audio signals and their labels. We verify that simple textual information from labels and advanced pretraining models enable more abundant semantic supervision for better performance. Specifically, we design a text encoder to capture the semantic information from the text extension of labels. Then we map the audio signals to align with the semantics of corresponding class labels via an audio encoder and a similarity calculation module so as to enforce the semantic consistency. Extensive experiments on two audio datasets, ESC-50 and US8K demonstrate that our proposed method consistently outperforms the compared audio classification methods.

cs.SD↗
arXiv2023

Spectral extrema of graphs with bounded clique number and matching number

For a set of graphs $\mathcal{F}$, let $\ex(n,\mathcal{F})$ and $\spex(n,\mathcal{F})$ denote the maximum number of edges and the maximum spectral radius of an $n$-vertex $\mathcal{F}$-free graph, respectively. Nikiforov ({\em LAA}, 2007) gave the spectral version of the Turán Theorem by showing that $\spex(n, K_{k+1})=λ(T_{k}(n))$, where $T_k(n)$ is the $k$-partite Turán graph on $n$ vertices. In the same year, Feng, Yu and Zhang ({\em LAA}) determined the exact value of $\spex(n, M_{s+1})$, where $M_{s+1}$ is a matching with $s+1$ edges. Recently, Alon and Frankl~(arXiv2210.15076) gave the exact value of $\ex(n,\{K_{k+1},M_{s+1}\})$. In this article, we give the spectral version of the result of Alon and Frankl by determining the exact value of $\spex(n,\{K_{k+1},M_{s+1}\})$ when $n$ is large.

math.CO↗
arXiv2022

Generically sharp decay for quasilinear wave equations with null condition

We are interested in the three-dimensional quasilinear wave equations with null condition. Global existence and pointwise decay for this model have been proved in the celebrated works of Klainerman \cite{Klainerman86} and Christodoulou \cite{Christodoulou86} for small smooth initial data. In this work, we illustrate the precise pointwise asymptotic behavior of the solutions for initial data posed on a hyperboloid and show that the decay rate $v^{-1}u^{-1}$ is optimal for a generic set of initial data.

math.AP↗