Search arXiv⌕ Search

arXiv subjects

Kun Zhou

Publications and source records attributed to Kun Zhou.

At least 289 records · Page 16Linked to original sources
arXiv2020

Towards High-Fidelity 3D Face Reconstruction from In-the-Wild Images Using Graph Convolutional Networks

3D Morphable Model (3DMM) based methods have achieved great success in recovering 3D face shapes from single-view images. However, the facial textures recovered by such methods lack the fidelity as exhibited in the input images. Recent work demonstrates high-quality facial texture recovering with generative networks trained from a large-scale database of high-resolution UV maps of face textures, which is hard to prepare and not publicly available. In this paper, we introduce a method to reconstruct 3D facial shapes with high-fidelity textures from single-view images in-the-wild, without the need to capture a large-scale face texture database. The main idea is to refine the initial texture generated by a 3DMM based method with facial details from the input image. To this end, we propose to use graph convolutional networks to reconstruct the detailed colors for the mesh vertices instead of reconstructing the UV map. Experiments show that our method can generate high-quality results and outperforms state-of-the-art methods in both qualitative and quantitative comparisons.

cs.CV↗
arXiv2020

Improving Conversational Recommender Systems via Knowledge Graph based Semantic Fusion

Conversational recommender systems (CRS) aim to recommend high-quality items to users through interactive conversations. Although several efforts have been made for CRS, two major issues still remain to be solved. First, the conversation data itself lacks of sufficient contextual information for accurately understanding users' preference. Second, there is a semantic gap between natural language expression and item-level user preference. To address these issues, we incorporate both word-oriented and entity-oriented knowledge graphs (KG) to enhance the data representations in CRSs, and adopt Mutual Information Maximization to align the word-level and entity-level semantic spaces. Based on the aligned semantic representations, we further develop a KG-enhanced recommender component for making accurate recommendations, and a KG-enhanced dialog component that can generate informative keywords or entities in the response text. Extensive experiments have demonstrated the effectiveness of our approach in yielding better performance on both recommendation and conversation tasks.

cs.CL↗
arXiv2020

AutoSweep: Recovering 3D Editable Objectsfrom a Single Photograph

This paper presents a fully automatic framework for extracting editable 3D objects directly from a single photograph. Unlike previous methods which recover either depth maps, point clouds, or mesh surfaces, we aim to recover 3D objects with semantic parts and can be directly edited. We base our work on the assumption that most human-made objects are constituted by parts and these parts can be well represented by generalized primitives. Our work makes an attempt towards recovering two types of primitive-shaped objects, namely, generalized cuboids and generalized cylinders. To this end, we build a novel instance-aware segmentation network for accurate part separation. Our GeoNet outputs a set of smooth part-level masks labeled as profiles and bodies. Then in a key stage, we simultaneously identify profile-body relations and recover 3D parts by sweeping the recognized profile along their body contour and jointly optimize the geometry to align with the recovered masks. Qualitative and quantitative experiments show that our algorithm can recover high quality 3D models and outperforms existing methods in both instance segmentation and 3D reconstruction. The dataset and code of AutoSweep are available at https://chenxin.tech/AutoSweep.html.

cs.CV↗
arXiv2020

SMART: Skeletal Motion Action Recognition aTtack

Adversarial attack has inspired great interest in computer vision, by showing that classification-based solutions are prone to imperceptible attack in many tasks. In this paper, we propose a method, SMART, to attack action recognizers which rely on 3D skeletal motions. Our method involves an innovative perceptual loss which ensures the imperceptibility of the attack. Empirical studies demonstrate that SMART is effective in both white-box and black-box scenarios. Its generalizability is evidenced on a variety of action recognizers and datasets. Its versatility is shown in different attacking strategies. Its deceitfulness is proven in extensive perceptual studies. Finally, SMART shows that adversarial attack on 3D skeletal motion, one type of time-series data, is significantly different from traditional adversarial attack problems.

cs.CV↗
arXiv2020

2D Black Phosphorus Carbide: Rippling and Formation of Nanotubes

The allotropes of a new layered material, phosphorus carbide (PC), have been predicted recently and a few of these predicted structures have already been successfully fabricated. Herein, by using first-principles calculations we investigated the effects of rippling a PC monolayer, one of the most stable modifications of layered PC, under large compressive strains. Similar to phosphorene, layered PC was found to have the extraordinary ability to bend and form ripples with large curvatures under a sufficiently large strain applied along its armchair direction. The band gap size, workfunction, and Young's modulus of rippled PC monolayer are predicted to be highly tunable by strain engineering. Moreover, a direct-indirect band gap transition is observed under the compressive strains in a range from 6 to 11%. Another important feature of PC monolayer rippled along the armchair direction is the possibility of its rolling to a PC nanotube (PCNT) under extreme compressive strains. These tubes of different sizes exhibit high thermal stability, possess a comparably high Young's modulus, and a well tunable band gap which can vary from 0 to 0.95 eV. In addition, for both structures, rippled PC and PCNTs, we have explained the changes in their properties under compressive strain in terms of the modification of their structural parameters.

cond-mat.mtrl-sci↗
arXiv2020

Improving Multi-Turn Response Selection Models with Complementary Last-Utterance Selection by Instance Weighting

Open-domain retrieval-based dialogue systems require a considerable amount of training data to learn their parameters. However, in practice, the negative samples of training data are usually selected from an unannotated conversation data set at random. The generated training data is likely to contain noise and affect the performance of the response selection models. To address this difficulty, we consider utilizing the underlying correlation in the data resource itself to derive different kinds of supervision signals and reduce the influence of noisy data. More specially, we consider a main-complementary task pair. The main task (\ie our focus) selects the correct response given the last utterance and context, and the complementary task selects the last utterance given the response and context. The key point is that the output of the complementary task is used to set instance weights for the main task. We conduct extensive experiments in two public datasets and obtain significant improvement in both datasets. We also investigate the variant of our approach in multiple aspects, and the results have verified the effectiveness of our approach.

cs.CL↗
arXiv2019

EmbedMask: Embedding Coupling for One-stage Instance Segmentation

Current instance segmentation methods can be categorized into segmentation-based methods that segment first then do clustering, and proposal-based methods that detect first then predict masks for each instance proposal using repooling. In this work, we propose a one-stage method, named EmbedMask, that unifies both methods by taking advantages of them. Like proposal-based methods, EmbedMask builds on top of detection models making it strong in detection capability. Meanwhile, EmbedMask applies extra embedding modules to generate embeddings for pixels and proposals, where pixel embeddings are guided by proposal embeddings if they belong to the same instance. Through this embedding coupling process, pixels are assigned to the mask of the proposal if their embeddings are similar. The pixel-level clustering enables EmbedMask to generate high-resolution masks without missing details from repooling, and the existence of proposal embedding simplifies and strengthens the clustering procedure to achieve high speed with higher performance than segmentation-based methods. Without any bells and whistles, EmbedMask achieves comparable performance as Mask R-CNN, which is the representative two-stage method, and can produce more detailed masks at a higher speed. Code is available at github.com/yinghdb/EmbedMask.

cs.CV↗
arXiv2019

Metastable Interlayer Frenkel Pair Defects by Dipole-like Strain Fields for Dimensional Distortion in Black Phosphorus

The low formation energy of atomic vacancies in black phosphorus allows it to serve as an ideal prototypical system for exploring the dynamics of interlayer interstitial-vacancy (I-V) pairs (i.e. Frenkel defects) which account for Wigner energy release. Based on a few-layer model of black phosphorus, we conduct discrete geometry analysis and investigate the structural dynamics of intimate interlayer Frenkel pairs from first-principles calculations. We reveal a highly metastable I-V pair state driven by anisotropic dipole-like strain fields which can build strong connections between neighbouring layers. In the 2D limit (monolayer), the intimate I-V pair exhibits a relatively low formation energy of 1.54 eV and is energetically favoured over its isolated constituents by up to 1.68 eV. The barrier for annihilation of the Frenkel pair is 1.46 eV in the bilayer, which is remarkably higher than that of similar defects in graphite. The findings reported in this work suggest that there exist rich bridging pathways in black phosphorus, leading to stable dimensional reduction and structural condensation on exposure to moderate electron excitation or thermal annealing. This study paves the way for creating novel dimensional-hybrid polymorphs of phosphorus via the introduction of such metastable interlayer I-V pair defects.

cond-mat.mtrl-sci↗
arXiv2019

Strain-driven superplasticity and modulation of electronic properties of ultrathin tin (II) oxide: A first-principles study

2D-layered tin (II) oxide (SnO) has recently emerged as a promising bipolar channel material for thin-film transistors and complementary metal-oxide-semiconductor devices. In this work, we present a first-principles investigation of the mechanical properties of ultrathin SnO, as well as the electronic implications of tensile strain ($ε$) under both uniaxial and biaxial conditions. Bulk-to-monolayer transition is found to significantly lower the Young's and shear moduli of SnO, highlighting the importance of interlayer Sn-Sn bonds in preserving structural integrity. Unprecedentedly, few-layer SnO exhibits superplasticity under uniaxial deformation conditions, with a critical strain to failure of up to 74% in the monolayer. Such superplastic behavior is ascribed to the formation of a tri-coordinated intermediate (referred to here as h-SnO) beyond $ε$ = 14%, which resembles a partially-recovered orthorhombic phase with relatively large work function and wide indirect band gap. The broad structural range of tin (II) oxide under strongly anisotropic mechanical loading suggests intriguing possibilities for realizing novel hybrid nanostructures of SnO through strain engineering. The findings reported in this study reveal fundamental insights into the mechanical behavior and strain-driven electronic properties of tin (II) oxide, opening up exciting avenues for the development of SnO-based nanoelectronic devices with new, non-conventional functionalities.

cond-mat.mtrl-sci↗
arXiv2019

Unsupervised Context Rewriting for Open Domain Conversation

Context modeling has a pivotal role in open domain conversation. Existing works either use heuristic methods or jointly learn context modeling and response generation with an encoder-decoder framework. This paper proposes an explicit context rewriting method, which rewrites the last utterance by considering context history. We leverage pseudo-parallel data and elaborate a context rewriting network, which is built upon the CopyNet with the reinforcement learning method. The rewritten utterance is beneficial to candidate retrieval, explainable context modeling, as well as enabling to employ a single-turn framework to the multi-turn scenario. The empirical results show that our model outperforms baselines in terms of the rewriting quality, the multi-turn response generation, and the end-to-end retrieval-based chatbots.

cs.CL↗
arXiv2019

HEMlets Pose: Learning Part-Centric Heatmap Triplets for Accurate 3D Human Pose Estimation

Estimating 3D human pose from a single image is a challenging task. This work attempts to address the uncertainty of lifting the detected 2D joints to the 3D space by introducing an intermediate state - Part-Centric Heatmap Triplets (HEMlets), which shortens the gap between the 2D observation and the 3D interpretation. The HEMlets utilize three joint-heatmaps to represent the relative depth information of the end-joints for each skeletal body part. In our approach, a Convolutional Network (ConvNet) is first trained to predict HEMlests from the input image, followed by a volumetric joint-heatmap regression. We leverage on the integral operation to extract the joint locations from the volumetric heatmaps, guaranteeing end-to-end learning. Despite the simplicity of the network design, the quantitative comparisons show a significant performance improvement over the best-of-grade method (by 20% on Human3.6M). The proposed method naturally supports training with "in-the-wild" images, where only weakly-annotated relative depth information of skeletal joints is available. This further improves the generalization ability of our model, as validated by qualitative comparisons on outdoor images.

cs.CV↗
arXiv2019

DeepSketchHair: Deep Sketch-based 3D Hair Modeling

We present sketchhair, a deep learning based tool for interactive modeling of 3D hair from 2D sketches. Given a 3D bust model as reference, our sketching system takes as input a user-drawn sketch (consisting of hair contour and a few strokes indicating the hair growing direction within a hair region), and automatically generates a 3D hair model, which matches the input sketch both globally and locally. The key enablers of our system are two carefully designed neural networks, namely, S2ONet, which converts an input sketch to a dense 2D hair orientation field; and O2VNet, which maps the 2D orientation field to a 3D vector field. Our system also supports hair editing with additional sketches in new views. This is enabled by another deep neural network, V2VNet, which updates the 3D vector field with respect to the new sketches. All the three networks are trained with synthetic data generated from a 3D hairstyle database. We demonstrate the effectiveness and expressiveness of our tool using a variety of hairstyles and also compare our method with prior art.

cs.GR↗
arXiv2019

Honest confidence sets for high-dimensional regression by projection and shrinkage

The issue of honesty in constructing confidence sets arises in nonparametric regression. While optimal rate in nonparametric estimation can be achieved and utilized to construct sharp confidence sets, severe degradation of confidence level often happens after estimating the degree of smoothness. Similarly, for high-dimensional regression, oracle inequalities for sparse estimators could be utilized to construct sharp confidence sets. Yet the degree of sparsity itself is unknown and needs to be estimated, causing the honesty problem. To resolve this issue, we develop a novel method to construct honest confidence sets for sparse high-dimensional linear regression. The key idea in our construction is to separate signals into a strong and a weak group, and then construct confidence sets for each group separately. This is achieved by a projection and shrinkage approach, the latter implemented via Stein estimation and the associated Stein unbiased risk estimate. Our confidence set is honest over the full parameter space without any sparsity constraints, while its diameter adapts to the optimal rate of $n^{-1/4}$ when the true parameter is indeed sparse. Through extensive numerical comparisons, we demonstrate that our method outperforms other competitors with big margins for finite samples, including oracle methods built upon the true sparsity of the underlying model.

stat.ME↗
arXiv2019

Ultrafast diffusive cross-sheet motion of lithium through antimonene with a 2+1 dimensional kinetics

Layered two-dimensional materials like graphene are highly appealing for lithium battery applications owing to their high surface-volume ratios. However, a critical issue that limits their practical applications is the confined motion of lithium atoms within their van der Waal's gaps, which is the leading cause for battery failure due to severe clustering and phase separation. Here we demonstrate that antimonene, an exfoliatable 2D material with a high structural stability, exhibits a highly mobile cross-sheet motion owing to its unique structural features. The advent of the vertically permeable channels opens a new pathway of lithium besides the normal motion along the basal plane, rendering a 2+1 dimensional kinetics. Specifically, our first-principles calculations combined with the discrete geometry analysis revealed that the energy barrier for a lithium atom to diffuse across the antimonene sheet is as low as 0.36 eV, which can be further reduced to 0.18 eV under a tensile strain of 4%. These ultralow diffusion barriers across the sheet can open a new dimension for controlling the motion of lithium atoms, leading to a new paradigm for high-performance lithium batteries or inorganic solid-state lithium-ion conductors.

cond-mat.mtrl-sci↗
arXiv2018

Adversarial 3D Human Pose Estimation via Multimodal Depth Supervision

In this paper, a novel deep-learning based framework is proposed to infer 3D human poses from a single image. Specifically, a two-phase approach is developed. We firstly utilize a generator with two branches for the extraction of explicit and implicit depth information respectively. During the training process, an adversarial scheme is also employed to further improve the performance. The implicit and explicit depth information with the estimated 2D joints generated by a widely used estimator, in the second step, are together fed into a deep 3D pose regressor for the final pose generation. Our method achieves MPJPE of 58.68mm on the ECCV2018 3D Human Pose Estimation Challenge.

cs.CV↗
arXiv2018

CaricatureShop: Personalized and Photorealistic Caricature Sketching

In this paper, we propose the first sketching system for interactively personalized and photorealistic face caricaturing. Input an image of a human face, the users can create caricature photos by manipulating its facial feature curves. Our system firstly performs exaggeration on the recovered 3D face model according to the edited sketches, which is conducted by assigning the laplacian of each vertex a scaling factor. To construct the mapping between 2D sketches and a vertex-wise scaling field, a novel deep learning architecture is developed. With the obtained 3D caricature model, two images are generated, one obtained by applying 2D warping guided by the underlying 3D mesh deformation and the other obtained by re-rendering the deformed 3D textured model. These two images are then seamlessly integrated to produce our final output. Due to the severely stretching of meshes, the rendered texture is of blurry appearances. A deep learning approach is exploited to infer the missing details for enhancing these blurry regions. Moreover, a relighting operation is invented to further improve the photorealism of the result. Both quantitative and qualitative experiment results validated the efficiency of our sketching system and the superiority of our proposed techniques against existing methods.

cs.CV↗