arXiv subjects

Xinghao Chen

Publications and source records attributed to Xinghao Chen.

At least 109 records · Page 6Linked to original sources

arXiv2021

OO Dra: an Algol-type binary formed through an extremely helium-poor mass accretion revealed by asteroseismology

Based on the two-minutes TESS data, we analyzed intrinsic oscillations of the primary component and identified seven confident independent $δ$ Scuti frequencies ($f_1$, $f_2$, $f_3$, $f_4$, $f_7$, $f_{11}$, and $f_{12}$). Both of single-star evolutionary models and mass-accreting models are computed to reproduce the $δ$ Scuti freqiencies. Fitting results of them match well with each other. The stellar parameters of the primary star yielded by asteroseismology are $M$ = $1.92^{+0.10}_{-0.02}$ $M_{\odot}$, $Z$ = 0.011$^{+0.006}_{-0.001}$, $R$ = $2.068^{+0.050}_{-0.007}$ $R_{\odot}$, $\log g$ = $4.090^{+0.010}_{-0.002}$, $T_{\rm eff}$ = $8346^{+244}_{-320}$ K, $L$ = $18.65^{+3.31}_{-2.82}$ $L_{\odot}$, which match well with the dynamic ones by the binary model. Furthermore, our asteroseismic results show that OO Dra is another Algol system that has just undergone the rapid mass-transfer stage. Fitting results of single-star evlutionary models indicate that the pulsator is helium-poor star with an age of 8.22$^{+0.12}_{-1.33}$ Myr, and the further mass-accreting models show that the primary star looks like an almost unevolved star formed by an extremely helium-poor mass accretion in Case A evolutionary scenario.

astro-ph.SR↗

arXiv2021

Winograd Algorithm for AdderNet

Adder neural network (AdderNet) is a new kind of deep model that replaces the original massive multiplications in convolutions by additions while preserving the high performance. Since the hardware complexity of additions is much lower than that of multiplications, the overall energy consumption is thus reduced significantly. To further optimize the hardware overhead of using AdderNet, this paper studies the winograd algorithm, which is a widely used fast algorithm for accelerating convolution and saving the computational costs. Unfortunately, the conventional Winograd algorithm cannot be directly applied to AdderNets since the distributive law in multiplication is not valid for the l1-norm. Therefore, we replace the element-wise multiplication in the Winograd equation by additions and then develop a new set of transform matrixes that can enhance the representation ability of output features to maintain the performance. Moreover, we propose the l2-to-l1 training strategy to mitigate the negative impacts caused by formal inconsistency. Experimental results on both FPGA and benchmarks show that the new method can further reduce the energy consumption without affecting the accuracy of the original AdderNet.

cs.LG↗

arXiv2021

Distilling Object Detectors via Decoupled Features

Knowledge distillation is a widely used paradigm for inheriting information from a complicated teacher network to a compact student network and maintaining the strong performance. Different from image classification, object detectors are much more sophisticated with multiple loss functions in which features that semantic information rely on are tangled. In this paper, we point out that the information of features derived from regions excluding objects are also essential for distilling the student detector, which is usually ignored in existing approaches. In addition, we elucidate that features from different regions should be assigned with different importance during distillation. To this end, we present a novel distillation algorithm via decoupled features (DeFeat) for learning a better student detector. Specifically, two levels of decoupled features will be processed for embedding useful information into the student, i.e., decoupled features from neck and decoupled proposals from classification head. Extensive experiments on various detectors with different backbones show that the proposed DeFeat is able to surpass the state-of-the-art distillation methods for object detection. For example, DeFeat improves ResNet50 based Faster R-CNN from 37.4% to 40.9% mAP, and improves ResNet50 based RetinaNet from 36.5% to 39.7% mAP on COCO benchmark. Our implementation is available at https://github.com/ggjy/DeFeat.pytorch.

cs.CV↗

arXiv2020

HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens

Neural Architecture Search (NAS) refers to automatically design the architecture. We propose an hourglass-inspired approach (HourNAS) for this problem that is motivated by the fact that the effects of the architecture often proceed from the vital few blocks. Acting like the narrow neck of an hourglass, vital blocks in the guaranteed path from the input to the output of a deep neural network restrict the information flow and influence the network accuracy. The other blocks occupy the major volume of the network and determine the overall network complexity, corresponding to the bulbs of an hourglass. To achieve an extremely fast NAS while preserving the high accuracy, we propose to identify the vital blocks and make them the priority in the architecture search. The search space of those non-vital blocks is further shrunk to only cover the candidates that are affordable under the computational resource constraints. Experimental results on the ImageNet show that only using 3 hours (0.1 days) with one GPU, our HourNAS can search an architecture that achieves a 77.0% Top-1 accuracy, which outperforms the state-of-the-art methods.

cs.CV↗

arXiv2020

VEGA: Towards an End-to-End Configurable AutoML Pipeline

Automated Machine Learning (AutoML) is an important industrial solution for automatic discovery and deployment of the machine learning models. However, designing an integrated AutoML system faces four great challenges of configurability, scalability, integrability, and platform diversity. In this work, we present VEGA, an efficient and comprehensive AutoML framework that is compatible and optimized for multiple hardware platforms. a) The VEGA pipeline integrates various modules of AutoML, including Neural Architecture Search (NAS), Hyperparameter Optimization (HPO), Auto Data Augmentation, Model Compression, and Fully Train. b) To support a variety of search algorithms and tasks, we design a novel fine-grained search space and its description language to enable easy adaptation to different search algorithms and tasks. c) We abstract the common components of deep learning frameworks into a unified interface. VEGA can be executed with multiple back-ends and hardwares. Extensive benchmark experiments on multiple tasks demonstrate that VEGA can improve the existing AutoML algorithms and discover new high-performance models against SOTA methods, e.g. the searched DNet model zoo for Ascend 10x faster than EfficientNet-B5 and 9.2x faster than RegNetX-32GF on ImageNet. VEGA is open-sourced at https://github.com/huawei-noah/vega.

cs.CV↗

arXiv2020

Kernel Based Progressive Distillation for Adder Neural Networks

Adder Neural Networks (ANNs) which only contain additions bring us a new way of developing deep neural networks with low energy consumption. Unfortunately, there is an accuracy drop when replacing all convolution filters by adder filters. The main reason here is the optimization difficulty of ANNs using $\ell_1$-norm, in which the estimation of gradient in back propagation is inaccurate. In this paper, we present a novel method for further improving the performance of ANNs without increasing the trainable parameters via a progressive kernel based knowledge distillation (PKKD) method. A convolutional neural network (CNN) with the same architecture is simultaneously initialized and trained as a teacher network, features and weights of ANN and CNN will be transformed to a new space to eliminate the accuracy drop. The similarity is conducted in a higher-dimensional space to disentangle the difference of their distributions using a kernel based method. Finally, the desired ANN is learned based on the information from both the ground-truth and teacher, progressively. The effectiveness of the proposed method for learning ANN with higher performance is then well-verified on several benchmarks. For instance, the ANN-50 trained using the proposed PKKD method obtains a 76.8\% top-1 accuracy on ImageNet dataset, which is 0.6\% higher than that of the ResNet-50.

cs.CV↗

arXiv2020

Optical Flow Distillation: Towards Efficient and Stable Video Style Transfer

Video style transfer techniques inspire many exciting applications on mobile devices. However, their efficiency and stability are still far from satisfactory. To boost the transfer stability across frames, optical flow is widely adopted, despite its high computational complexity, e.g. occupying over 97% inference time. This paper proposes to learn a lightweight video style transfer network via knowledge distillation paradigm. We adopt two teacher networks, one of which takes optical flow during inference while the other does not. The output difference between these two teacher networks highlights the improvements made by optical flow, which is then adopted to distill the target student network. Furthermore, a low-rank distillation loss is employed to stabilize the output of student network by mimicking the rank of input videos. Extensive experiments demonstrate that our student network without an optical flow module is still able to generate stable video and runs much faster than the teacher network.

cs.CV↗

arXiv2020

KIC 10736223: An Algol-type eclipsing binary just undergone the rapid mass-transfer stage

This paper reports the discovery of an Algol system KIC 10736223 that just past the rapid mass transfer stage. From the light curve and radial-velocity modelling we find KIC 10736223 to be a detached Algol system with the less-massive secondary nearly filling its Roche lobe. Based on the short-cadence Kepler data, we analyzed intrinsic oscillations of the pulsator and identified six secured independent $δ$ Scuti-type pulsation modes ($f_{1}$, $f_3$, $f_{9}$, $f_{19}$, $f_{42}$, and $f_{48}$). We compute two grids of theoretical models to reproduce the $δ$ Scuti freqiencies, and find fitting results of mass-accreting models meet well with those of single-star evolutionary models. The fundamental parameters of the primary star yielded with asteroseismology are $M$ = $1.57^{+0.05}_{-0.09}$ $M_{\odot}$, $Z$ = 0.009 $\pm$ 0.001, $R$ = $1.484^{+0.016}_{-0.028}$ $R_{\odot}$, $\log g$ = $4.291^{+0.004}_{-0.009}$, $T_{\rm eff}$ = $7748^{+230}_{-378}$ K, $L$ = $7.136^{+1.014}_{-1.519}$ $L_{\odot}$. The asteroseismic parameters match well with the dynamical parameters derived from the binary model. Moreover, our asteroseismic results show that the pulsator is an almost unevolved star with an age between 9.46-11.65 Myr for single-star evolutionary models and 2.67-3.14 Myr for mass-accreting models. Thereofore, KIC 10736223 may be an Algol system that has just undergone the rapid mass-transfer process.

astro-ph.SR↗

arXiv2020

Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection

Neural Architecture Search (NAS) has achieved great success in image classification task. Some recent works have managed to explore the automatic design of efficient backbone or feature fusion layer for object detection. However, these methods focus on searching only one certain component of object detector while leaving others manually designed. We identify the inconsistency between searched component and manually designed ones would withhold the detector of stronger performance. To this end, we propose a hierarchical trinity search framework to simultaneously discover efficient architectures for all components (i.e. backbone, neck, and head) of object detector in an end-to-end manner. In addition, we empirically reveal that different parts of the detector prefer different operators. Motivated by this, we employ a novel scheme to automatically screen different sub search spaces for different components so as to perform the end-to-end search for each component on the corresponding sub search space efficiently. Without bells and whistles, our searched architecture, namely Hit-Detector, achieves 41.4\% mAP on COCO minival set with 27M parameters. Our implementation is available at https://github.com/ggjy/HitDet.pytorch.

cs.CV↗

arXiv2020

CARS: Continuous Evolution for Efficient Neural Architecture Search

Searching techniques in most of existing neural architecture search (NAS) algorithms are mainly dominated by differentiable methods for the efficiency reason. In contrast, we develop an efficient continuous evolutionary approach for searching neural networks. Architectures in the population that share parameters within one SuperNet in the latest generation will be tuned over the training dataset with a few epochs. The searching in the next evolution generation will directly inherit both the SuperNet and the population, which accelerates the optimal network generation. The non-dominated sorting strategy is further applied to preserve only results on the Pareto front for accurately updating the SuperNet. Several neural networks with different model sizes and performances will be produced after the continuous search with only 0.4 GPU days. As a result, our framework provides a series of networks with the number of parameters ranging from 3.7M to 5.1M under mobile settings. These networks surpass those produced by the state-of-the-art methods on the benchmark ImageNet dataset.

cs.CV↗

arXiv2019

Exploring the convective core of the hybrid $δ$ Scuti-$γ$ Doradus star CoRoT 100866999 with asteroseismology

We computed a grid of theoretical models to fit the $δ$ Scuti frequencies of CoRoT 100866999 detected earlier from the CoRoT timeserials. The pulsating primary star is determined to be a main sequence star with a rotation period of $4.1^{+0.6}_{-0.5}$ days, rotating slower than the orbital motion. The fundamental parameters of the primary star are determined to be $M$ = $1.71^{+0.13}_{-0.04}$ $M_{\odot}$, $Z=0.012^{+0.004}_{-0.000}$, $f_{\rm ov}$ = $0.02^{+0.00}_{-0.02}$, $T_{\rm eff}$ = $8024^{+249}_{-297}$ K, $L$ = $11.898^{+2.156}_{-1.847}$ $L_{\odot}$, $\log g$ = $4.166^{+0.013}_{-0.002}$, $R$ = $1.787^{+0.040}_{-0.016}$ $R_{\odot}$, and $X_{\rm c}$ = 0.488$^{+0.011}_{-0.020}$, matching well those obtained from the eclipsing light curve analysis. Based on the model fittings, $p_1$ and $p_5$ are suggested to be two dipole modes, and $p_3$, $p_4$, $p_6$, and $p_7$ to be four quadrupole modes. In particular, $p_4$ and $p_7$ are identified as two components of one quintuplet. Based on the best-fitting model, we find that $p_1$ is a g mode and the other nonradial modes have pronounced mixed characters, which give strong constraints on the convective core. Finally, the relative size of the convective core of CoRoT 100866999 is determined to $R_{\rm conv}/R$ = $0.0931^{+0.0003}_{-0.0013}$.

astro-ph.SR↗

arXiv2019

Bi-stream Pose Guided Region Ensemble Network for Fingertip Localization from Stereo Images

In human-computer interaction, it is important to accurately estimate the hand pose especially fingertips. However, traditional approaches for fingertip localization mainly rely on depth images and thus suffer considerably from the noise and missing values. Instead of depth images, stereo images can also provide 3D information of hands and promote 3D hand pose estimation. There are nevertheless limitations on the dataset size, global viewpoints, hand articulations and hand shapes in the publicly available stereo-based hand pose datasets. To mitigate these limitations and promote further research on hand pose estimation from stereo images, we propose a new large-scale binocular hand pose dataset called THU-Bi-Hand, offering a new perspective for fingertip localization. In the THU-Bi-Hand dataset, there are 447k pairs of stereo images of different hand shapes from 10 subjects with accurate 3D location annotations of the wrist and five fingertips. Captured with minimal restriction on the range of hand motion, the dataset covers large global viewpoint space and hand articulation space. To better present the performance of fingertip localization on THU-Bi-Hand, we propose a novel scheme termed Bi-stream Pose Guided Region Ensemble Network (Bi-Pose-REN). It extracts more representative feature regions around joint points in the feature maps under the guidance of the previously estimated pose. The feature regions are integrated hierarchically according to the topology of hand joints to regress the refined hand pose. Bi-Pose-REN and several existing methods are evaluated on THU-Bi-Hand so that benchmarks are provided for further research. Experimental results show that our new method has achieved the best performance on THU-Bi-Hand.

cs.CV↗

arXiv2019

Asteroseismology of the pre-main-sequence $δ$ Scuti pulsator IP Persei

A grid of theoretical models are computed to fit the 9 oscillation modes of IP Per detected earlier from a multi-site ground-based campaign. Fitting results show that there are two sets of theoretical models that could reproduce the observed frequencies more or less euqally well. In view of other available spectroscopic and photometric measurements, our best fitting stellar parameters for IP Per are $\upsilon_{\rm e} = 91^{+5}_{-3}$ km s$^{-1}$, $Z=0.009^{+0.004}_{-0.001}$, $M = 1.64^{+0.10}_{-0.04}$ $M_{\odot}$, $T_{\rm eff} = 7766^{+348}_{-173}$ K, $\log L/L_{\odot} = 1.125^{+0.094}_{-0.046}$, $\log g = 4.041^{+0.008}_{-0.003}$ dex, $R = 2.022^{+0.042}_{-0.018}$ $R_{\odot}$, $τ_{0} = 8711^{+68}_{-35}$ s, age = $7.39^{+0.76}_{-0.46}$ Myr. Meanwhile, IP Per is found to be a pre-main sequence star where CN cycle has not yet reached the equilibrium state. At present fourteen percent of C12 have been turned into N14. Based on the best-fitting model, $f_{6}$ is identified as a radial mode, $f_{1}$ and $f_{2}$ as two dipole modes, and $f_{3}$, $f_{4}$, $f_{5}$, $f_{7}$, $f_{8}$, and $f_{9}$ as six quadrupole modes.

astro-ph.SR↗

arXiv2018

Determining the evolutionary stage of the delta Scuti star HIP 80088 with asteroseismology

We have computed a grid of theoretical models to fit the twelve oscillation modes of HIP 80088 observed by K2. HIP 80088 is determined to be a pre-main sequence star, in which CN cycle has not arrived at the equilibrium state. Mass fractions of C12 and N14 in metal composition are 0.1277$^{+0.0064}_{-0.0049}$ and 0.1092$^{+0.0057}_{-0.0074}$ respectively, indicating that twenty-eight percent of C12 have turned into N14. Meanwhile, our fitting results show that physical parameters of HIP 80088 converge to a small range: $M$ = 1.68$-$1.78 $M _{\odot}$, $Z$ = 0.015$-$0.018, $\upsilon_{\rm e}$ = 120$-$136 km s$^{-1}$, $\log g$ = 4.114$-$4.125, $R$ = 1.882$-$1.914 $R_{\odot}$, $τ_{0}$ = 7636$-$7723 s, and age = 9.03$-$10.21 Myr. Based on our model fittings, $f_{3}$ is suggested to be one radial mode, $f_{2}$, $f_{4}$, $f_{8}$, and $f_{11}$ to be four $\ell$ = 1 modes, and $f_{1}$, $f_{5}$, $f_{6}$, $f_{7}$, $f_{9}$, $f_{10}$, and $f_{12}$ to be seven $\ell$ = 2 modes. In particularly, we find that ($f_{2}$, $f_{4}$, $f_{8}$) form one complete triplet with the averaged frequency spacing of 16.045 $μ$Hz , and ($f_{5}$, $f_{7}$, $f_{9}$, $f_{10}$) form four components of one quintuplet with the averaged frequency spacing of 13.388 $μ$Hz. The two averaged frequency spacings are not equal. Based on the best-fitting model, those $\ell$ = 2 modes of HIP 80088 are found to be mixed modes, which are p-dominated modes with pronounced g-mode features, while oscillation modes with $\ell$ = 1 are p modes.

astro-ph.SR↗

arXiv2018

Interactive Hand Pose Estimation: Boosting accuracy in localizing extended finger joints

Accurate 3D hand pose estimation plays an important role in Human Machine Interaction (HMI). In the reality of HMI, joints in fingers stretching out, especially corresponding fingertips, are much more important than other joints. We propose a novel method to refine stretching-out finger joint locations after obtaining rough hand pose estimation. It first detects which fingers are stretching out, then neighbor pixels of certain joint vote for its new location based on random forests. The algorithm is tested on two public datasets: MSRA15 and ICVL. After the refinement stage of stretching-out fingers, errors of predicted HMI finger joint locations are significantly reduced. Mean error of all fingertips reduces around 5mm (relatively more than 20%). Stretching-out fingertip locations are even more precise, which in MSRA15 reduces 10.51mm (relatively 41.4%).

cs.CV↗

arXiv2018

Pose Guided Structured Region Ensemble Network for Cascaded Hand Pose Estimation

Hand pose estimation from a single depth image is an essential topic in computer vision and human computer interaction. Despite recent advancements in this area promoted by convolutional neural network, accurate hand pose estimation is still a challenging problem. In this paper we propose a Pose guided structured Region Ensemble Network (Pose-REN) to boost the performance of hand pose estimation. The proposed method extracts regions from the feature maps of convolutional neural network under the guide of an initially estimated pose, generating more optimal and representative features for hand pose estimation. The extracted feature regions are then integrated hierarchically according to the topology of hand joints by employing tree-structured fully connections. A refined estimation of hand pose is directly regressed by the proposed network and the final hand pose is obtained by utilizing an iterative cascaded method. Comprehensive experiments on public hand pose datasets demonstrate that our proposed method outperforms state-of-the-art algorithms.

cs.CV↗

arXiv2018

Two-Stream Binocular Network: Accurate Near Field Finger Detection Based On Binocular Images

Fingertip detection plays an important role in human computer interaction. Previous works transform binocular images into depth images. Then depth-based hand pose estimation methods are used to predict 3D positions of fingertips. Different from previous works, we propose a new framework, named Two-Stream Binocular Network (TSBnet) to detect fingertips from binocular images directly. TSBnet first shares convolutional layers for low level features of right and left images. Then it extracts high level features in two-stream convolutional networks separately. Further, we add a new layer: binocular distance measurement layer to improve performance of our model. To verify our scheme, we build a binocular hand image dataset, containing about 117k pairs of images in training set and 10k pairs of images in test set. Our methods achieve an average error of 10.9mm on our test set, outperforming previous work by 5.9mm (relatively 35.1%).

cs.CV↗

arXiv2018

Depth-Based 3D Hand Pose Estimation: From Current Achievements to Future Goals

In this paper, we strive to answer two questions: What is the current state of 3D hand pose estimation from depth images? And, what are the next challenges that need to be tackled? Following the successful Hands In the Million Challenge (HIM2017), we investigate the top 10 state-of-the-art methods on three tasks: single frame 3D pose estimation, 3D hand tracking, and hand pose estimation during object interaction. We analyze the performance of different CNN structures with regard to hand shape, joint visibility, view point and articulation distributions. Our findings include: (1) isolated 3D hand pose estimation achieves low mean errors (10 mm) in the view point range of [70, 120] degrees, but it is far from being solved for extreme view points; (2) 3D volumetric representations outperform 2D CNNs, better capturing the spatial structure of the depth data; (3) Discriminative methods still generalize poorly to unseen hand shapes; (4) While joint occlusions pose a challenge for most methods, explicit modeling of structure constraints can significantly narrow the gap between errors on visible and occluded joints.

cs.CV↗