arXiv subjects

Zihan Zhou

Publications and source records attributed to Zihan Zhou.

At least 145 records · Page 8Linked to original sources

arXiv2022

Deep Learning Assisted Optimization for 3D Reconstruction from Single 2D Line Drawings

In this paper, we revisit the long-standing problem of automatic reconstruction of 3D objects from single line drawings. Previous optimization-based methods can generate compact and accurate 3D models, but their success rates depend heavily on the ability to (i) identifying a sufficient set of true geometric constraints, and (ii) choosing a good initial value for the numerical optimization. In view of these challenges, we propose to train deep neural networks to detect pairwise relationships among geometric entities (i.e., edges) in the 3D object, and to predict initial depth value of the vertices. Our experiments on a large dataset of CAD models show that, by leveraging deep learning in a geometric constraint solving pipeline, the success rate of optimization-based 3D reconstruction can be significantly improved.

cs.CV↗

arXiv2022

One Small Step for an Inflaton, One Giant Leap for Inflation: a novel non-Gaussian tail and primordial black holes

We report a novel prediction from single-field inflation that even a tiny step in the inflaton potential can change our perception of primordial non-Gaussianities of the curvature perturbation. Our analysis focuses on the tail of probability distribution generated by an upward step transition between two stages of slow-roll evolution. The nontrivial background dynamics with off-attractor behavior is identified. By using a non-perturbative $δN$ analysis, we explicitly show that a highly non-Gaussian tail can be generated by a tiny upward step, even when the conventional nonlinearity parameters $f_{NL}$, $g_{NL}$, etc. remain small. With this example, we demonstrate for the first time the sensitive dependence of non-perturbative effects on the tail of probability distribution. Our scenario has an inconceivable application to primordial black holes by either significantly boosting their abundance or completely forbidding their appearance.

astro-ph.CO↗

arXiv2022

Birefringence Tomography for Axion Cloud

An axion cloud surrounding a supermassive black hole can be naturally produced through the superradiance process. Its existence can be examined by the axion induced birefringence effect. It predicts an oscillation of the electric vector position angle of linearly polarized radiations. Stringent constraints of the existence of the axion in a particular mass window has been obtained based on the recent Event Horizon Telescope measurement on M87$^\star$. The future Very-Long-Baseline Interferometry (VLBI) observations will be able to measure the vicinity of many supermassive black holes, thus it opens the possibility to search for the existence of axions in a wide mass regime. In this paper, we study how different black hole properties and accretion flows influence the signatures of the axion induced birefringence. We include the impacts of black hole inclination angles, spins, magnetic fields, plasma velocity distributions, the thickness of the accretion flows. We pay special attention to characterize the washout effects induced by the finite thickness of the accretion flows and the lensed photons. Based on this study, we give prospects on how to optimize the axion search using future VLBI observations, such as the next-generation Event Horizon Telescope, to further increase the sensitivity.

hep-ph↗

arXiv2022

Vanishing of black hole tidal Love numbers from scattering amplitudes

We extract the black hole (BH) static tidal deformability coefficients (Love numbers) and their spin-0 and spin-1 analogs by comparing on-shell amplitudes for fields to scatter off a spinning BH in the worldline effective field theory (EFT) and in general relativity (GR). We point out that the GR amplitudes due to tidal effects originate entirely from the BH potential region. Thus, they can be separated from gravitational non-linearities in the wave region, whose proper treatment requires higher order EFT loop calculations. In particular, the elastic scattering in the near field approximation is produced exclusively by tidal effects. We find this contribution to vanish identically, which implies that the static Love numbers of Kerr BHs are zero for all types of perturbations. We also reproduce the known behavior of scalar Love numbers for higher dimensional BHs. Our results are manifestly gauge-invariant and coordinate-independent, thereby providing a valuable consistency check for the commonly used off-shell methods.

hep-th↗

arXiv2022

End-to-end Graph-constrained Vectorized Floorplan Generation with Panoptic Refinement

The automatic generation of floorplans given user inputs has great potential in architectural design and has recently been explored in the computer vision community. However, the majority of existing methods synthesize floorplans in the format of rasterized images, which are difficult to edit or customize. In this paper, we aim to synthesize floorplans as sequences of 1-D vectors, which eases user interaction and design customization. To generate high fidelity vectorized floorplans, we propose a novel two-stage framework, including a draft stage and a multi-round refining stage. In the first stage, we encode the room connectivity graph input by users with a graph convolutional network (GCN), then apply an autoregressive transformer network to generate an initial floorplan sequence. To polish the initial design and generate more visually appealing floorplans, we further propose a novel panoptic refinement network(PRN) composed of a GCN and a transformer network. The PRN takes the initial generated sequence as input and refines the floorplan design while encouraging the correct room connectivity with our proposed geometric loss. We have conducted extensive experiments on a real-world floorplan dataset, and the results show that our method achieves state-of-the-art performance under different settings and evaluation metrics.

cs.CV↗

arXiv2022

Model-based Synthetic Data-driven Learning (MOST-DL): Application in Single-shot T2 Mapping with Severe Head Motion Using Overlapping-echo Acquisition

Use of synthetic data has provided a potential solution for addressing unavailable or insufficient training samples in deep learning-based magnetic resonance imaging (MRI). However, the challenge brought by domain gap between synthetic and real data is usually encountered, especially under complex experimental conditions. In this study, by combining Bloch simulation and general MRI models, we propose a framework for addressing the lack of training data in supervised learning scenarios, termed MOST-DL. A challenging application is demonstrated to verify the proposed framework and achieve motion-robust T2 mapping using single-shot overlapping-echo acquisition. We decompose the process into two main steps: (1) calibrationless parallel reconstruction for ultra-fast pulse sequence and (2) intra-shot motion correction for T2 mapping. To bridge the domain gap, realistic textures from a public database and various imperfection simulations were explored. The neural network was first trained with pure synthetic data and then evaluated with in vivo human brain. Both simulation and in vivo experiments show that the MOST-DL method significantly reduces ghosting and motion artifacts in T2 maps in the presence of unpredictable subject movement and has the potential to be applied to motion-prone patients in the clinic.

eess.IV↗

arXiv2022

Continuously Discovering Novel Strategies via Reward-Switching Policy Optimization

We present Reward-Switching Policy Optimization (RSPO), a paradigm to discover diverse strategies in complex RL environments by iteratively finding novel policies that are both locally optimal and sufficiently different from existing ones. To encourage the learning policy to consistently converge towards a previously undiscovered local optimum, RSPO switches between extrinsic and intrinsic rewards via a trajectory-based novelty measurement during the optimization process. When a sampled trajectory is sufficiently distinct, RSPO performs standard policy optimization with extrinsic rewards. For trajectories with high likelihood under existing policies, RSPO utilizes an intrinsic diversity reward to promote exploration. Experiments show that RSPO is able to discover a wide spectrum of strategies in a variety of domains, ranging from single-agent particle-world tasks and MuJoCo continuous control to multi-agent stag-hunt games and StarCraftII challenges.

cs.LG↗

arXiv2022

Deep Depth from Focus with Differential Focus Volume

Depth-from-focus (DFF) is a technique that infers depth using the focus change of a camera. In this work, we propose a convolutional neural network (CNN) to find the best-focused pixels in a focal stack and infer depth from the focus estimation. The key innovation of the network is the novel deep differential focus volume (DFV). By computing the first-order derivative with the stacked features over different focal distances, DFV is able to capture both the focus and context information for focus analysis. Besides, we also introduce a probability regression mechanism for focus estimation to handle sparsely sampled focal stacks and provide uncertainty estimation to the final prediction. Comprehensive experiments demonstrate that the proposed model achieves state-of-the-art performance on multiple datasets with good generalizability and fast speed.

cs.CV↗

arXiv2022

Neural Face Identification in a 2D Wireframe Projection of a Manifold Object

In computer-aided design (CAD) systems, 2D line drawings are commonly used to illustrate 3D object designs. To reconstruct the 3D models depicted by a single 2D line drawing, an important key is finding the edge loops in the line drawing which correspond to the actual faces of the 3D object. In this paper, we approach the classical problem of face identification from a novel data-driven point of view. We cast it as a sequence generation problem: starting from an arbitrary edge, we adopt a variant of the popular Transformer model to predict the edges associated with the same face in a natural order. This allows us to avoid searching the space of all possible edge loops with various hand-crafted rules and heuristics as most existing methods do, deal with challenging cases such as curved surfaces and nested edge loops, and leverage additional cues such as face types. We further discuss how possibly imperfect predictions can be used for 3D object reconstruction.

cs.CV↗

arXiv2021

Lerna: Transformer Architectures for Configuring Error Correction Tools for Short- and Long-Read Genome Sequencing

Sequencing technologies are prone to errors, making error correction (EC) necessary for downstream applications. EC tools need to be manually configured for optimal performance. We find that the optimal parameters (e.g., k-mer size) are both tool- and dataset-dependent. Moreover, evaluating the performance (i.e., Alignment-rate or Gain) of a given tool usually relies on a reference genome, but quality reference genomes are not always available. We introduce Lerna for the automated configuration of k-mer-based EC tools. Lerna first creates a language model (LM) of the uncorrected genomic reads; then, calculates the perplexity metric to evaluate the corrected reads for different parameter choices. Next, it finds the one that produces the highest alignment rate without using a reference genome. The fundamental intuition of our approach is that the perplexity metric is inversely correlated with the quality of the assembly after error correction. Results: First, we show that the best k-mer value can vary for different datasets, even for the same EC tool. Second, we show the gains of our LM using its component attention-based transformers. We show the model's estimation of the perplexity metric before and after error correction. The lower the perplexity after correction, the better the k-mer size. We also show that the alignment rate and assembly quality computed for the corrected reads are strongly negatively correlated with the perplexity, enabling the automated selection of k-mer values for better error correction, and hence, improved assembly quality. Additionally, we show that our attention-based models have significant runtime improvement for the entire pipeline -- 18X faster than previous works, due to parallelizing the attention mechanism and the use of JIT compilation for GPU inferencing.

q-bio.GN↗

arXiv2021

Beating the Lyth bound by parametric resonance during inflation

We propose a novel mechanism for enhancing the primordial gravitational waves without significantly affecting the curvature perturbations produced during inflation. This is achieved due to non-linear sourcing of resonantly amplified scalar field fluctuations. Our result is an explicit scale-dependent counter-example of the famous Lyth bound, which opens up a promising perspective of producing detectable inflationary tensor modes with low-scale inflation and a sub-Planckian field excursion. We explicitly demonstrate the testability of our mechanism with upcoming Cosmic Microwave Background B-mode observations.

astro-ph.CO↗

arXiv2021

Temporal Induced Self-Play for Stochastic Bayesian Games

One practical requirement in solving dynamic games is to ensure that the players play well from any decision point onward. To satisfy this requirement, existing efforts focus on equilibrium refinement, but the scalability and applicability of existing techniques are limited. In this paper, we propose Temporal-Induced Self-Play (TISP), a novel reinforcement learning-based framework to find strategies with decent performances from any decision point onward. TISP uses belief-space representation, backward induction, policy learning, and non-parametric approximation. Building upon TISP, we design a policy-gradient-based algorithm TISP-PG. We prove that TISP-based algorithms can find approximate Perfect Bayesian Equilibrium in zero-sum one-sided stochastic Bayesian games with finite horizon. We test TISP-based algorithms in various games, including finitely repeated security games and a grid-world game. The results show that TISP-PG is more scalable than existing mathematical programming-based methods and significantly outperforms other learning-based methods.

cs.MA↗

arXiv2021

Towards Robust Human Trajectory Prediction in Raw Videos

Human trajectory prediction has received increased attention lately due to its importance in applications such as autonomous vehicles and indoor robots. However, most existing methods make predictions based on human-labeled trajectories and ignore the errors and noises in detection and tracking. In this paper, we study the problem of human trajectory forecasting in raw videos, and show that the prediction accuracy can be severely affected by various types of tracking errors. Accordingly, we propose a simple yet effective strategy to correct the tracking failures by enforcing prediction consistency over time. The proposed "re-tracking" algorithm can be applied to any existing tracking and prediction pipelines. Experiments on public benchmark datasets demonstrate that the proposed method can improve both tracking and prediction performance in challenging real-world scenarios. The code and data are available at https://git.io/retracking-prediction.

cs.CV↗

arXiv2021

Optimized multi-axis spiral projection MR fingerprinting with subspace reconstruction for rapid whole-brain high-isotropic-resolution quantitative imaging

Purpose: To improve image quality and accelerate the acquisition of 3D MRF. Methods: Building on the multi-axis spiral-projection MRF technique, a subspace reconstruction with locally low rank (LLR) constraint and a modified spiral-projection spatiotemporal encoding scheme termed tiny-golden-angle-shuffling (TGAS) were implemented for rapid whole-brain high-resolution quantitative mapping. The LLR regularization parameter and the number of subspace bases were tuned using retrospective in-vivo data and simulated examinations, respectively. B0 inhomogeneity correction using multi-frequency interpolation was incorporated into the subspace reconstruction to further improve the image quality by mitigating blurring caused by off-resonance effect. Results: The proposed MRF acquisition and reconstruction framework can produce provide high quality 1-mm isotropic whole-brain quantitative maps in a total acquisition time of 1 minute 55 seconds, with higher-quality results than ones obtained from the previous approach in 6 minutes. The comparison of quantitative results indicates that neither the subspace reconstruction nor the TGAS trajectory induce bias for T1 and T2 mapping. High quality whole-brain MRF data were also obtained at 0.66-mm isotropic resolution in 4 minutes using the proposed technique, where the increased resolution was shown to improve visualization of subtle brain structures. Conclusion: The proposed TGAS-SPI-MRF with optimized spiral-projection trajectory and subspace reconstruction can enable high-resolution quantitative mapping with faster acquisition speed.

physics.med-ph↗

arXiv2020

Probing new physics with multi-vacua quantum tunnelings beyond standard model through gravitational waves

We report on a novel phenomenon of particle cosmology, which features specific cosmological phase transitions via quantum tunnelings through multiple vacua. The latter is inspired by the axiverse ideas and enables us to probe the associated new physics models through a potential observation of specific patterns in the stochastic gravitational waves background. Multiple vacua may induce the nucleation of co-existing bubbles over the phase transition epoch, hence enhancing the overall process of bubbles' nucleation. Our detailed analysis of semi-analytical and numerical solutions to the bounce equations of the path integral in three vacua case has enabled us to determine the existence of three instanton solutions. This new mechanism of cosmological phase transitions clearly predicts a possibly sizeable new source of gravitational waves, with its energy spectrum being featured with particular patterns, which could be probed by the future gravitational wave interferometers.

astro-ph.CO↗

arXiv2020

Primordial black holes and gravitational waves from resonant amplification during inflation

We present a new realization of the resonant production of primordial black holes as well as gravitational waves in a two-stage inflation model consisting of a scalar field ϕwith an axion-monodromy-like periodic structure in the potential that governs the first stage and another field χwith a hilltop-like potential that dominates the second stage. The parametric resonance seeded by the periodic structure at the first stage amplifies the perturbations of both fields inside the Hubble radius. While the evolution of the background trajectory experiences a turn as the oscillatory barrier height increases, the amplified perturbations of χremain as they are and contribute to the final curvature perturbation. It turns out that the primordial power spectrum displays a significant resonant peak on small scales, which can lead to an abundant production of primordial black holes. Furthermore, gravitational waves are also generated from the resonantly enhanced field perturbations during inflation, the amplitude of which may be constrained by future gravitational wave interferometers.

astro-ph.CO↗

arXiv2020

Data-Driven Distributed State Estimation and Behavior Modeling in Sensor Networks

Nowadays, the prevalence of sensor networks has enabled tracking of the states of dynamic objects for a wide spectrum of applications from autonomous driving to environmental monitoring and urban planning. However, tracking real-world objects often faces two key challenges: First, due to the limitation of individual sensors, state estimation needs to be solved in a collaborative and distributed manner. Second, the objects' movement behavior is unknown, and needs to be learned using sensor observations. In this work, for the first time, we formally formulate the problem of simultaneous state estimation and behavior learning in a sensor network. We then propose a simple yet effective solution to this new problem by extending the Gaussian process-based Bayes filters (GP-BayesFilters) to an online, distributed setting. The effectiveness of the proposed method is evaluated on tracking objects with unknown movement behaviors using both synthetic data and data collected from a multi-robot platform.

cs.RO↗

arXiv2020

Structured3D: A Large Photo-realistic Dataset for Structured 3D Modeling

Recently, there has been growing interest in developing learning-based methods to detect and utilize salient semi-global or global structures, such as junctions, lines, planes, cuboids, smooth surfaces, and all types of symmetries, for 3D scene modeling and understanding. However, the ground truth annotations are often obtained via human labor, which is particularly challenging and inefficient for such tasks due to the large number of 3D structure instances (e.g., line segments) and other factors such as viewpoints and occlusions. In this paper, we present a new synthetic dataset, Structured3D, with the aim of providing large-scale photo-realistic images with rich 3D structure annotations for a wide spectrum of structured 3D modeling tasks. We take advantage of the availability of professional interior designs and automatically extract 3D structures from them. We generate high-quality images with an industry-leading rendering engine. We use our synthetic dataset in combination with real images to train deep networks for room layout estimation and demonstrate improved performance on benchmark datasets.

cs.CV↗