arXiv subjects

Tianqi Li

Publications and source records attributed to Tianqi Li.

35 records · Page 2Linked to original sources

arXiv2024

Learning Dynamic Tetrahedra for High-Quality Talking Head Synthesis

Recent works in implicit representations, such as Neural Radiance Fields (NeRF), have advanced the generation of realistic and animatable head avatars from video sequences. These implicit methods are still confronted by visual artifacts and jitters, since the lack of explicit geometric constraints poses a fundamental challenge in accurately modeling complex facial deformations. In this paper, we introduce Dynamic Tetrahedra (DynTet), a novel hybrid representation that encodes explicit dynamic meshes by neural networks to ensure geometric consistency across various motions and viewpoints. DynTet is parameterized by the coordinate-based networks which learn signed distance, deformation, and material texture, anchoring the training data into a predefined tetrahedra grid. Leveraging Marching Tetrahedra, DynTet efficiently decodes textured meshes with a consistent topology, enabling fast rendering through a differentiable rasterizer and supervision via a pixel loss. To enhance training efficiency, we incorporate classical 3D Morphable Models to facilitate geometry learning and define a canonical space for simplifying texture learning. These advantages are readily achievable owing to the effective geometric representation employed in DynTet. Compared with prior works, DynTet demonstrates significant improvements in fidelity, lip synchronization, and real-time performance according to various metrics. Beyond producing stable and visually appealing synthesis videos, our method also outputs the dynamic meshes which is promising to enable many emerging applications.

cs.CV↗

arXiv2023

Out-of-Distribution Detection in Long-Tailed Recognition with Calibrated Outlier Class Learning

Existing out-of-distribution (OOD) methods have shown great success on balanced datasets but become ineffective in long-tailed recognition (LTR) scenarios where 1) OOD samples are often wrongly classified into head classes and/or 2) tail-class samples are treated as OOD samples. To address these issues, current studies fit a prior distribution of auxiliary/pseudo OOD data to the long-tailed in-distribution (ID) data. However, it is difficult to obtain such an accurate prior distribution given the unknowingness of real OOD samples and heavy class imbalance in LTR. A straightforward solution to avoid the requirement of this prior is to learn an outlier class to encapsulate the OOD samples. The main challenge is then to tackle the aforementioned confusion between OOD samples and head/tail-class samples when learning the outlier class. To this end, we introduce a novel calibrated outlier class learning (COCL) approach, in which 1) a debiased large margin learning method is introduced in the outlier class learning to distinguish OOD samples from both head and tail classes in the representation space and 2) an outlier-class-aware logit calibration method is defined to enhance the long-tailed classification confidence. Extensive empirical results on three popular benchmarks CIFAR10-LT, CIFAR100-LT, and ImageNet-LT demonstrate that COCL substantially outperforms state-of-the-art OOD detection methods in LTR while being able to improve the classification accuracy on ID data. Code is available at https://github.com/mala-lab/COCL.

cs.CV↗

arXiv2023

Learning Adversarial Semantic Embeddings for Zero-Shot Recognition in Open Worlds

Zero-Shot Learning (ZSL) focuses on classifying samples of unseen classes with only their side semantic information presented during training. It cannot handle real-life, open-world scenarios where there are test samples of unknown classes for which neither samples (e.g., images) nor their side semantic information is known during training. Open-Set Recognition (OSR) is dedicated to addressing the unknown class issue, but existing OSR methods are not designed to model the semantic information of the unseen classes. To tackle this combined ZSL and OSR problem, we consider the case of "Zero-Shot Open-Set Recognition" (ZS-OSR), where a model is trained under the ZSL setting but it is required to accurately classify samples from the unseen classes while being able to reject samples from the unknown classes during inference. We perform large experiments on combining existing state-of-the-art ZSL and OSR models for the ZS-OSR task on four widely used datasets adapted from the ZSL task, and reveal that ZS-OSR is a non-trivial task as the simply combined solutions perform badly in distinguishing the unseen-class and unknown-class samples. We further introduce a novel approach specifically designed for ZS-OSR, in which our model learns to generate adversarial semantic embeddings of the unknown classes to train an unknowns-informed ZS-OSR classifier. Extensive empirical results show that our method 1) substantially outperforms the combined solutions in detecting the unknown classes while retaining the classification accuracy on the unseen classes and 2) achieves similar superiority under generalized ZS-OSR settings.

cs.CV↗

arXiv2023

Emergent Cooperative Behavior in Distributed Target Tracking with Unknown Occlusions

Tracking multiple moving objects of interest (OOI) with multiple robot systems (MRS) has been addressed by active sensing that maintains a shared belief of OOIs and plans the motion of robots to maximize the information quality. Mobility of robots enables the behavior of pursuing better visibility, which is constrained by sensor field of view (FoV) and occlusion objects. We first extend prior work to detect, maintain and share occlusion information explicitly, allowing us to generate occlusion-aware planning even if a priori semantic occlusion information is unavailable. The efficacy of active sensing approaches is often evaluated according to estimation error and information gain metrics. However, these metrics do not directly explain the level of cooperative behavior engendered by the active sensing algorithms. Next, we extract different emergent cooperative behaviors that stem from the same underlying algorithms but manifest differently under differing scenarios. In particular, we highlight and demonstrate three emergent behavior patterns in active sensing MRS: (i) Change of tracking responsibility between agents when tracking trajectories with divergent directions or due to a re-allocation of the resource among heterogeneous agents; (ii) Awareness of occlusions to a trajectory and temporal leave-and-return of the sensing agent; (iii) Sharing of local occlusion objects in MRS that subsequently improves the awareness of occlusion.

cs.RO↗

arXiv2022

Revealing intra-urban spatial structure through an exploratory analysis by combining road network abstraction model and taxi trajectory data

The unprecedented urbanization in China has dramatically changed the urban spatial structure of cities. With the proliferation of individual-level geospatial big data, previous studies have widely used the network abstraction model to reveal the underlying urban spatial structure. However, the construction of network abstraction models primarily focuses on the topology of the road network without considering individual travel flows along with the road networks. Individual travel flows reflect the urban dynamics, which can further help understand the underlying spatial structure. This study therefore aims to reveal the intra-urban spatial structure by integrating the road network abstraction model and individual travel flows. To achieve this goal, we 1) quantify the spatial interaction relatedness of road segments based on the Word2Vec model using large volumes of taxi trip data, then 2) characterize the road abstraction network model according to the identified spatial interaction relatedness, and 3) implement a community detection algorithm to reveal sub-regions of a city. Our results reveal three levels of hierarchical spatial structures in the Wuhan metropolitan area. This study provides a data-driven approach to the investigation of urban spatial structure via identifying traffic interaction patterns on the road network, offering insights to urban planning practice and transportation management.

cs.SI↗

arXiv2022

Probing the valence quark region of nucleons with Z bosons at LHCb

Because of its forward coverage, LHCb can probe the valence quark distributions of protons and nuclei at small and large Bjorken-x with high precision. This proceeding presents new LHCb measurements of Z boson production in association with charm jet in the forward region of proton-proton collisions and Z boson production in proton-lead collisions. Z+charm jet production could be sensitive to a valence-like intrinsic-charm component in the proton wave function. The measurements of Z production in proton-lead collisions provide new constraints on the partonic structure of nucleons inside nuclei. Comparisons between the results and calculations with parton distribution functions are also discussed.

hep-ex↗

arXiv2022

SMA-NBO: A Sequential Multi-Agent Planning with Nominal Belief-State Optimization in Target Tracking

In target tracking with mobile multi-sensor systems, sensor deployment impacts the observation capabilities and the resulting state estimation quality. Based on a partially observable Markov decision process (POMDP) formulation comprised of the observable sensor dynamics, unobservable target states, and accompanying observation laws, we present a distributed information-driven solution approach to the multi-agent target tracking problem, namely, sequential multi-agent nominal belief-state optimization (SMA-NBO). SMA-NBO seeks to minimize the expected tracking error via receding horizon control including a heuristic expected cost-to-go (HECTG). SMA-NBO incorporates a computationally efficient approximation of the target belief-state over the horizon. The agent-by-agent decision-making is capable of leveraging on-board (edge) compute for selecting (sub-optimal) target-tracking maneuvers exhibiting non-myopic cooperative fleet behavior. The optimization problem explicitly incorporates semantic information defining target occlusions from a world model. To illustrate the efficacy of our approach, a random occlusion forest environment is simulated. SMA-NBO is compared to other baseline approaches. The simulation results show SMA-NBO 1) maintains tracking performance and reduces the computational cost by replacing the calculation of the expected target trajectory with a single sample trajectory based on maximum a posteriori estimation; 2) generates cooperative fleet decision by sequentially optimizing single-agent policy with efficient usage of other agents' policy of intent; 3) aptly incorporates the multiple weighted trace penalty (MWTP) HECTG, which improves tracking performance with a computationally efficient heuristic.

cs.MA↗

arXiv2022

Optimizing Consensus-based Multi-target Tracking with Multiagent Rollout Control Policies

This paper considers a multiagent, connected, robotic fleet where the primary functionality of the agents is sensing. A distributed multi-sensor control strategy maximizes the value of the collective sensing capability of the fleet, using an information-driven approach. Each agent individually performs sensor processing (Kalman Filtering and Joint Probabilistic Data Association) to identify trajectories (and associated distributions). Using communications with its neighbors the agents enhance the prediction of the trajectories using a Consensus of Information approach that iteratively calculates the Kullback-Leibler average of trajectory distributions, enabling the calculation of the value of the collective information for the fleet. The dynamics of the agents, the evolution of the identified trajectories for each agent, and the dynamics of individual observed objects are captured as a Partially Observable Markov Decision Process (POMDP). Using this POMDP and applying rollout with receding horizon control, an optimized non-myopic control policy that maximizes the collective fleet information value is synthesized. Simulations are performed for a scenario with three heterogeneous UAVs performing coordinated target tracking that illustrate the proposed methodology and compare the centralized approach with a contemporary sequential multiagent distributed decision technique.

cs.MA↗

arXiv2021

Towards Adversarial Robustness via Transductive Learning

There has been emerging interest to use transductive learning for adversarial robustness (Goldwasser et al., NeurIPS 2020; Wu et al., ICML 2020). Compared to traditional "test-time" defenses, these defense mechanisms "dynamically retrain" the model based on test time input via transductive learning; and theoretically, attacking these defenses boils down to bilevel optimization, which seems to raise the difficulty for adaptive attacks. In this paper, we first formalize and analyze modeling aspects of transductive robustness. Then, we propose the principle of attacking model space for solving bilevel attack objectives, and present an instantiation of the principle which breaks previous transductive defenses. These attacks thus point to significant difficulties in the use of transductive learning to improve adversarial robustness. To this end, we present new theoretical and empirical evidence in support of the utility of transductive learning.

cs.LG↗

arXiv2020

An efficient algorithm for $1$-dimensional (persistent) path homology

This paper focuses on developing an efficient algorithm for analyzing a directed network (graph) from a topological viewpoint. A prevalent technique for such topological analysis involves computation of homology groups and their persistence. These concepts are well suited for spaces that are not directed. As a result, one needs a concept of homology that accommodates orientations in input space. Path-homology developed for directed graphs by Grigor'yan, Lin, Muranov and Yau has been effectively adapted for this purpose recently by Chowdhury and Mémoli. They also give an algorithm to compute this path-homology. Our main contribution in this paper is an algorithm that computes this path-homology and its persistence more efficiently for the $1$-dimensional ($H_1$) case. In developing such an algorithm, we discover various structures and their efficient computations that aid computing the $1$-dimensional path-homnology. We implement our algorithm and present some preliminary experimental results.

cs.CG↗

arXiv2018

Improving Multi-Person Pose Estimation using Label Correction

Significant attention is being paid to multi-person pose estimation methods recently, as there has been rapid progress in the field owing to convolutional neural networks. Especially, recent method which exploits part confidence maps and Part Affinity Fields (PAFs) has achieved accurate real-time prediction of multi-person keypoints. However, human annotated labels are sometimes inappropriate for learning models. For example, if there is a limb that extends outside an image, a keypoint for the limb may not have annotations because it exists outside of the image, and thus the labels for the limb can not be generated. If a model is trained with data including such missing labels, the output of the model for the location, even though it is correct, is penalized as a false positive, which is likely to cause negative effects on the performance of the model. In this paper, we point out the existence of some patterns of inappropriate labels, and propose a novel method for correcting such labels with a teacher model trained on such incomplete data. Experiments on the COCO dataset show that training with the corrected labels improves the performance of the model and also speeds up training.

cs.CV↗

arXiv2018

Full-body High-resolution Anime Generation with Progressive Structure-conditional Generative Adversarial Networks

We propose Progressive Structure-conditional Generative Adversarial Networks (PSGAN), a new framework that can generate full-body and high-resolution character images based on structural information. Recent progress in generative adversarial networks with progressive training has made it possible to generate high-resolution images. However, existing approaches have limitations in achieving both high image quality and structural consistency at the same time. Our method tackles the limitations by progressively increasing the resolution of both generated images and structural conditions during training. In this paper, we empirically demonstrate the effectiveness of this method by showing the comparison with existing approaches and video generation results of diverse anime characters at 1024x1024 based on target pose sequences. We also create a novel dataset containing full-body 1024x1024 high-resolution images and exact 2D pose keypoints using Unity 3D Avatar models.

cs.CV↗

arXiv2018

Efficient algorithms for computing a minimal homology basis

Efficient computation of shortest cycles which form a homology basis under $\mathbb{Z}_2$-additions in a given simplicial complex $\mathcal{K}$ has been researched actively in recent years. When the complex $\mathcal{K}$ is a weighted graph with $n$ vertices and $m$ edges, the problem of computing a shortest (homology) cycle basis is known to be solvable in $O(m^2n/\log n+ n^2m)$-time. Several works \cite{borradaile2017minimum, greedy} have addressed the case when the complex $\mathcal{K}$ is a $2$-manifold. The complexity of these algorithms depends on the rank $g$ of the one-dimensional homology group of $\mathcal{K}$. This rank $g$ has a lower bound of $Θ(n)$, where $n$ denotes the number of simplices in $\mathcal{K}$, giving an $O(n^4)$ worst-case time complexity for the algorithms in \cite{borradaile2017minimum,greedy}. This worst-case complexity is improved in \cite{annotation} to $O(n^ω+ n^2g^{ω-1})$ for general simplicial complexes where $ω< 2.3728639$ \cite{le2014powers} is the matrix multiplication exponent. Taking $g=Θ(n)$, this provides an $O(n^{ω+1})$ worst-case algorithm. In this paper, we improve this time complexity. Combining the divide and conquer technique from \cite{DivideConquer} with the use of annotations from \cite{annotation}, we present an algorithm that runs in $O(n^ω+n^2g)$ time giving the first $O(n^3)$ worst-case algorithm for general complexes. If instead of minimal basis, we settle for an approximate basis, we can improve the running time even further. We show that a $2$-approximate minimal homology basis can be computed in $O(n^ω\sqrt{n \log n})$ expected time. We also study more general measures for defining the minimal basis and identify reasonable conditions on these measures that allow computing a minimal basis efficiently.

math.AT↗

arXiv2014

Ultrafast Probes of Nonequilibrium Hole Spin Relaxation in Ferromagnetic Semiconductor GaMnAs

We directly measure the hole spin lifetime in ferromagnetic GaMnAs via time- and polarization-resolved spectroscopy. Below the Curie temperature Tc, an ultrafast photoexcitation with linearly-polarized light is shown to create a non-equilibrium hole spin population via the dynamical polarization of holes through p-d exchange scattering with ferromagnetically-ordered Mn spins, and we characterize their relaxation dynamics. The observed relaxation consists of a distinct three-step recovery : (i) femtosecond (fs) hole spin relaxation ~ $160-200 fs, (ii) picosecond (ps) hole energy relaxation ~ 1-2 ps, and (iii) a coherent, damped Mn spin precession with a period of ~ 250 ps. The transient amplitude of the hole spin component diminishes with increasing temperature, directly following the ferromagnetic order, while the hole energy amplitude shows negligible temperature change, consistent with our interpretation. Our results thus establish the hole spin lifetimes in ferromagnetic semiconductors and demonstrate a novel spectroscopy method for studying non-equilibrium hole spins in the presence of correlation and magnetic order.

cond-mat.mtrl-sci↗

arXiv2014

Ultrafast observation of critical nematic fluctuations and giant magnetoelastic coupling in iron pnictides

Many of the iron pnictides have strongly anisotropic normal-state characteristics, important for the exotic magnetic and superconducting behavior these materials exhibit. Yet, the origin of the observed anisotropy is unclear. Electronically driven nematicity has been suggested, but distinguishing this as an independent degree of freedom from magnetic and structural orders is difficult, as these couple together to break the same tetragonal symmetry. Here we use time-resolved polarimetry to reveal critical nematic fluctuations in unstrained Ba(Fe_(1-x)Co_x)_2As_2. The femtosecond anisotropic response, which arises from the two-fold in-plane anisotropy of the complex refractive index, displays a characteristic two-step recovery absent in the isotropic response. The fast recovery appears only in the magnetically ordered state, whereas the slow one persists in the paramagnetic phase with a critical divergence approaching the structural transition temperature. The dynamics also reveal a gigantic magnetoelastic coupling that far exceeds electron-spin and electron-phonon couplings, opposite to conventional magnetic metals.

cond-mat.str-el↗

arXiv2013

Transient Charge and Energy Balance in Graphene Induced by Ultrafast Photoexcitation

Ultrafast optical pump-probe spectroscopy measurement on monolayer graphene observes significant optical nonlinearities. We show that strongly photoexcited graphene monolayers with 35 fs pulses quasi-instantaneously build up a broadband, inverted Dirac fermion population. Optical gain emerges and directly manifests itself via a negative conductivity at the near-infrared region for the first 200fs, where stimulated emission completely compensates absorption loss in the graphene layer. To quantitatively investigate this transient, extremely dense photoexcited Dirac-fermion state, we construct a two-chemical-potential model, in addition to a time-dependent transient carrier temperature above lattice temperature, to describe the population inverted electronic state metastable on the time scale of tens of femtoseconds generated by a strong exciting pulse. The calculated transient optical conductivity reveals a complete bleaching of absorption, which sets the saturation density during the pulse propagation. Particularly, the model calculation reproduces the negative optical conductivity at lower frequencies in the states close to saturation, corroborating the observed femtosecond stimulated emission and optical gain in the wide near-infrared window.

cond-mat.mes-hall↗

arXiv2013

Post-transient relaxation in graphene after an intense laser pulse

High intensity laser pulses were recently shown to induce a population inverted transient state in graphene [T. Li et al. Phys. Rev. Lett. 108, 167401 (2012)]. Using a combination of hydrodynamic arguments and a kinetic theory we determine the post-transient state relaxation of hot, dense, population inverted electrons towards equilibrium. The cooling rate and charge-imbalance relaxation rate are determined from the Boltzmann-equation including electron-phonon scattering. We show that the relaxation of the population inversion, driven by inter-band scattering processes, is much slower than the relaxation of the electron temperature, which is determined by intra-band scattering processes. This insight may be of relevance for the application of graphene as an optical gain medium.

cond-mat.mes-hall↗