arXiv subjects

Zicheng Liu

Publications and source records attributed to Zicheng Liu.

At least 199 records · Page 11Linked to original sources

arXiv2021

TransMOT: Spatial-Temporal Graph Transformer for Multiple Object Tracking

Tracking multiple objects in videos relies on modeling the spatial-temporal interactions of the objects. In this paper, we propose a solution named TransMOT, which leverages powerful graph transformers to efficiently model the spatial and temporal interactions among the objects. TransMOT effectively models the interactions of a large number of objects by arranging the trajectories of the tracked objects as a set of sparse weighted graphs, and constructing a spatial graph transformer encoder layer, a temporal transformer encoder layer, and a spatial graph transformer decoder layer based on the graphs. TransMOT is not only more computationally efficient than the traditional Transformer, but it also achieves better tracking accuracy. To further improve the tracking speed and accuracy, we propose a cascade association framework to handle low-score detections and long-term occlusions that require large computational resources to model in TransMOT. The proposed method is evaluated on multiple benchmark datasets including MOT15, MOT16, MOT17, and MOT20, and it achieves state-of-the-art performance on all the datasets.

cs.CV↗

arXiv2021

Disentanglement-based Cross-Domain Feature Augmentation for Effective Unsupervised Domain Adaptive Person Re-identification

Unsupervised domain adaptive (UDA) person re-identification (ReID) aims to transfer the knowledge from the labeled source domain to the unlabeled target domain for person matching. One challenge is how to generate target domain samples with reliable labels for training. To address this problem, we propose a Disentanglement-based Cross-Domain Feature Augmentation (DCDFA) strategy, where the augmented features characterize well the target and source domain data distributions while inheriting reliable identity labels. Particularly, we disentangle each sample feature into a robust domain-invariant/shared feature and a domain-specific feature, and perform cross-domain feature recomposition to enhance the diversity of samples used in the training, with the constraints of cross-domain ReID loss and domain classification loss. Each recomposed feature, obtained based on the domain-invariant feature (which enables a reliable inheritance of identity) and an enhancement from a domain specific feature (which enables the approximation of real distributions), is thus an "ideal" augmentation. Extensive experimental results demonstrate the effectiveness of our method, which achieves the state-of-the-art performance.

cs.CV↗

arXiv2021

Revisiting Dynamic Convolution via Matrix Decomposition

Recent research in dynamic convolution shows substantial performance boost for efficient CNNs, due to the adaptive aggregation of K static convolution kernels. It has two limitations: (a) it increases the number of convolutional weights by K-times, and (b) the joint optimization of dynamic attention and static convolution kernels is challenging. In this paper, we revisit it from a new perspective of matrix decomposition and reveal the key issue is that dynamic convolution applies dynamic attention over channel groups after projecting into a higher dimensional latent space. To address this issue, we propose dynamic channel fusion to replace dynamic attention over channel groups. Dynamic channel fusion not only enables significant dimension reduction of the latent space, but also mitigates the joint optimization difficulty. As a result, our method is easier to train and requires significantly fewer parameters without sacrificing accuracy. Source code is at https://github.com/liyunsheng13/dcd.

cs.CV↗

arXiv2021

VIVO: Visual Vocabulary Pre-Training for Novel Object Captioning

It is highly desirable yet challenging to generate image captions that can describe novel objects which are unseen in caption-labeled training data, a capability that is evaluated in the novel object captioning challenge (nocaps). In this challenge, no additional image-caption training data, other thanCOCO Captions, is allowed for model training. Thus, conventional Vision-Language Pre-training (VLP) methods cannot be applied. This paper presents VIsual VOcabulary pretraining (VIVO) that performs pre-training in the absence of caption annotations. By breaking the dependency of paired image-caption training data in VLP, VIVO can leverage large amounts of paired image-tag data to learn a visual vocabulary. This is done by pre-training a multi-layer Transformer model that learns to align image-level tags with their corresponding image region features. To address the unordered nature of image tags, VIVO uses a Hungarian matching loss with masked tag prediction to conduct pre-training. We validate the effectiveness of VIVO by fine-tuning the pre-trained model for image captioning. In addition, we perform an analysis of the visual-text alignment inferred by our model. The results show that our model can not only generate fluent image captions that describe novel objects, but also identify the locations of these objects. Our single model has achieved new state-of-the-art results on nocaps and surpassed the human CIDEr score.

cs.CV↗

arXiv2020

3D Human motion anticipation and classification

Human motion prediction and understanding is a challenging problem. Due to the complex dynamic of human motion and the non-deterministic aspect of future prediction. We propose a novel sequence-to-sequence model for human motion prediction and feature learning, trained with a modified version of generative adversarial network, with a custom loss function that takes inspiration from human motion animation and can control the variation between multiple predicted motion from the same input poses. Our model learns to predict multiple future sequences of human poses from the same input sequence. We show that the discriminator learns general presentation of human motion by using the learned feature in action recognition task. Furthermore, to quantify the quality of the non-deterministic predictions, we simultaneously train a motion-quality-assessment network that learns the probability that a given sequence of poses is a real human motion or not. We test our model on two of the largest human pose datasets: NTURGB-D and Human3.6M. We train on both single and multiple action types. Its predictive power for motion estimation is demonstrated by generating multiple plausible futures from the same input and show the effect of each of the loss functions. Furthermore, we show that it takes less than half the number of epochs to train an activity recognition network by using the feature learned from the discriminator.

cs.CV↗

arXiv2020

MicroNet: Towards Image Recognition with Extremely Low FLOPs

In this paper, we present MicroNet, which is an efficient convolutional neural network using extremely low computational cost (e.g. 6 MFLOPs on ImageNet classification). Such a low cost network is highly desired on edge devices, yet usually suffers from a significant performance degradation. We handle the extremely low FLOPs based upon two design principles: (a) avoiding the reduction of network width by lowering the node connectivity, and (b) compensating for the reduction of network depth by introducing more complex non-linearity per layer. Firstly, we propose Micro-Factorized convolution to factorize both pointwise and depthwise convolutions into low rank matrices for a good tradeoff between the number of channels and input/output connectivity. Secondly, we propose a new activation function, named Dynamic Shift-Max, to improve the non-linearity via maxing out multiple dynamic fusions between an input feature map and its circular channel shift. The fusions are dynamic as their parameters are adapted to the input. Building upon Micro-Factorized convolution and dynamic Shift-Max, a family of MicroNets achieve a significant performance gain over the state-of-the-art in the low FLOP regime. For instance, MicroNet-M1 achieves 61.1% top-1 accuracy on ImageNet classification with 12 MFLOPs, outperforming MobileNetV3 by 11.3%.

cs.CV↗

arXiv2020

Silicon substrate significantly alters dipole-dipole resolution in coherent microscope

Influences of a substrate below samples in imaging performances are studied by reaching the solution to the dyadic Green's function, where the substrate is modeled as half space in the sample region. Then, theoretical and numerical analysis are performed in terms of magnification, depth of field, and resolution. Various settings including positions of dipoles, the distance of the substrate to the focal plane and dipole polarization are considered. Methods to measure the resolution of $z$-polarized dipoles are also presented since the modified Rayleigh limit cannot be applied directly. The silicon substrate and the glass substrate are studied with a water immersion objective lens. The high contrast between silicon and water leads to significant disturbances on imaging.

physics.optics↗

arXiv2020

Dynamic ReLU

Rectified linear units (ReLU) are commonly used in deep neural networks. So far ReLU and its generalizations (non-parametric or parametric) are static, performing identically for all input samples. In this paper, we propose dynamic ReLU (DY-ReLU), a dynamic rectifier of which parameters are generated by a hyper function over all in-put elements. The key insight is that DY-ReLU encodes the global context into the hyper function, and adapts the piecewise linear activation function accordingly. Compared to its static counterpart, DY-ReLU has negligible extra computational cost, but significantly more representation capability, especially for light-weight neural networks. By simply using DY-ReLU for MobileNetV2, the top-1 accuracy on ImageNet classification is boosted from 72.0% to 76.2% with only 5% additional FLOPs.

cs.CV↗

arXiv2020

Cross-Domain Complementary Learning Using Pose for Multi-Person Part Segmentation

Supervised deep learning with pixel-wise training labels has great successes on multi-person part segmentation. However, data labeling at pixel-level is very expensive. To solve the problem, people have been exploring to use synthetic data to avoid the data labeling. Although it is easy to generate labels for synthetic data, the results are much worse compared to those using real data and manual labeling. The degradation of the performance is mainly due to the domain gap, i.e., the discrepancy of the pixel value statistics between real and synthetic data. In this paper, we observe that real and synthetic humans both have a skeleton (pose) representation. We found that the skeletons can effectively bridge the synthetic and real domains during the training. Our proposed approach takes advantage of the rich and realistic variations of the real data and the easily obtainable labels of the synthetic data to learn multi-person part segmentation on real images without any human-annotated labels. Through experiments, we show that without any human labeling, our method performs comparably to several state-of-the-art approaches which require human labeling on Pascal-Person-Parts and COCO-DensePose datasets. On the other hand, if part labels are also available in the real-images during training, our method outperforms the supervised state-of-the-art methods by a large margin. We further demonstrate the generalizability of our method on predicting novel keypoints in real images where no real data labels are available for the novel keypoints detection. Code and pre-trained models are available at https://github.com/kevinlin311tw/CDCL-human-part-segmentation

cs.CV↗

arXiv2020

Rethinking Classification and Localization for Object Detection

Two head structures (i.e. fully connected head and convolution head) have been widely used in R-CNN based detectors for classification and localization tasks. However, there is a lack of understanding of how does these two head structures work for these two tasks. To address this issue, we perform a thorough analysis and find an interesting fact that the two head structures have opposite preferences towards the two tasks. Specifically, the fully connected head (fc-head) is more suitable for the classification task, while the convolution head (conv-head) is more suitable for the localization task. Furthermore, we examine the output feature maps of both heads and find that fc-head has more spatial sensitivity than conv-head. Thus, fc-head has more capability to distinguish a complete object from part of an object, but is not robust to regress the whole object. Based upon these findings, we propose a Double-Head method, which has a fully connected head focusing on classification and a convolution head for bounding box regression. Without bells and whistles, our method gains +3.5 and +2.8 AP on MS COCO dataset from Feature Pyramid Network (FPN) baselines with ResNet-50 and ResNet-101 backbones, respectively.

cs.CV↗

arXiv2020

Dynamic Convolution: Attention over Convolution Kernels

Light-weight convolutional neural networks (CNNs) suffer performance degradation as their low computational budgets constrain both the depth (number of convolution layers) and the width (number of channels) of CNNs, resulting in limited representation capability. To address this issue, we present Dynamic Convolution, a new design that increases model complexity without increasing the network depth or width. Instead of using a single convolution kernel per layer, dynamic convolution aggregates multiple parallel convolution kernels dynamically based upon their attentions, which are input dependent. Assembling multiple kernels is not only computationally efficient due to the small kernel size, but also has more representation power since these kernels are aggregated in a non-linear way via attention. By simply using dynamic convolution for the state-of-the-art architecture MobileNetV3-Small, the top-1 accuracy of ImageNet classification is boosted by 2.9% with only 4% additional FLOPs and 2.9 AP gain is achieved on COCO keypoint detection.

cs.CV↗

arXiv2020

Learning Nonparametric Human Mesh Reconstruction from a Single Image without Ground Truth Meshes

Nonparametric approaches have shown promising results on reconstructing 3D human mesh from a single monocular image. Unlike previous approaches that use a parametric human model like skinned multi-person linear model (SMPL), and attempt to regress the model parameters, nonparametric approaches relax the heavy reliance on the parametric space. However, existing nonparametric methods require ground truth meshes as their regression target for each vertex, and obtaining ground truth mesh labels is very expensive. In this paper, we propose a novel approach to learn human mesh reconstruction without any ground truth meshes. This is made possible by introducing two new terms into the loss function of a graph convolutional neural network (Graph CNN). The first term is the Laplacian prior that acts as a regularizer on the reconstructed mesh. The second term is the part segmentation loss that forces the projected region of the reconstructed mesh to match the part segmentation. Experimental results on multiple public datasets show that without using 3D ground truth meshes, the proposed approach outperforms the previous state-of-the-art approaches that require ground truth meshes for training.

cs.CV↗

arXiv2020

Discrepancies of measured SAR between traditional and fast measuring systems

Human exposure to mobile devices is traditionally measured by a system in which the human body (or head) is modelled by a phantom and the energy absorbed from the device is estimated based on the electric fields measured with a single probe. Such a system suffers from low efficiency due to repeated volumetric scanning within the phantom needed to capture the absorbed energy throughout the volume. To speed up the measurement, fast SAR (specific absorption rate) measuring systems have been developed. However, discrepancies of measured results are observed between traditional and fast measuring systems. In this paper, the discrepancies in terms of post-processing procedures after the measurement of electric field (or its amplitude) are investigated. Here, the concerned fast measuring system estimates SAR based on the reconstructed field of the region of interest while the amplitude and phase of electric field are measured on a single plane with a probe array. The numerical results presented indicate that the fast SAR measuring system has the potential to yield more accurate estimations than the traditional system, but no conclusion can be made on which kind of system is superior without knowledge of the field-reconstruction algorithms and the emitting source.

eess.SP↗

arXiv2019

Surrogate modeling of indoor down-link human exposure based on sparse polynomial chaos expansion

Human exposure induced by wireless communication systems increasingly draws the public attention. Here, an indoor down-link scenario is concerned and the exposure level is statistically analyzed. The electromagnetic field (EMF) emitted by a WiFi box is measured and electromagnetic dosimetry features are evaluated from the whole-body specific absorption rate as computed with a Finite-Difference Time-Domain (a.k.a. FDTD) code. Due to computational cost, a statistical analysis is performed based on a surrogate model, which is constructed by means of so-called sparse polynomial chaos expansion (PCE), where the inner cross validation (ICV) is used to select the optimal hyperparameters during the model construction and assess the model performance. However, the ICV error is optimized and the model assessment tends to be overly optimistic with small data sets. The method of cross-model validation is used and outer cross validation is carried out for the model assessment. The effects of the data preprocessing are investigated as well. Based on the surrogate model, the global sensitivity of the exposure to input parameters is analyzed from Sobol' indices.

stat.AP↗

arXiv2019

Large Scale Incremental Learning

Modern machine learning suffers from catastrophic forgetting when learning new classes incrementally. The performance dramatically degrades due to the missing data of old classes. Incremental learning methods have been proposed to retain the knowledge acquired from the old classes, by using knowledge distilling and keeping a few exemplars from the old classes. However, these methods struggle to scale up to a large number of classes. We believe this is because of the combination of two factors: (a) the data imbalance between the old and new classes, and (b) the increasing number of visually similar classes. Distinguishing between an increasing number of visually similar classes is particularly challenging, when the training data is unbalanced. We propose a simple and effective method to address this data imbalance issue. We found that the last fully connected layer has a strong bias towards the new classes, and this bias can be corrected by a linear model. With two bias parameters, our method performs remarkably well on two large datasets: ImageNet (1000 classes) and MS-Celeb-1M (10000 classes), outperforming the state-of-the-art algorithms by 11.1% and 13.2% respectively.

cs.CV↗

arXiv2018

Reinforced Temporal Attention and Split-Rate Transfer for Depth-Based Person Re-Identification

We address the problem of person re-identification from commodity depth sensors. One challenge for depth-based recognition is data scarcity. Our first contribution addresses this problem by introducing split-rate RGB-to-Depth transfer, which leverages large RGB datasets more effectively than popular fine-tuning approaches. Our transfer scheme is based on the observation that the model parameters at the bottom layers of a deep convolutional neural network can be directly shared between RGB and depth data while the remaining layers need to be fine-tuned rapidly. Our second contribution enhances re-identification for video by implementing temporal attention as a Bernoulli-Sigmoid unit acting upon frame-level features. Since this unit is stochastic, the temporal attention parameters are trained using reinforcement learning. Extensive experiments validate the accuracy of our method in person re-identification from depth sequences. Finally, in a scenario where subjects wear unseen clothes, we show large performance gains compared to a state-of-the-art model which relies on RGB data.

cs.CV↗

arXiv2018

Incremental Classifier Learning with Generative Adversarial Networks

In this paper, we address the incremental classifier learning problem, which suffers from catastrophic forgetting. The main reason for catastrophic forgetting is that the past data are not available during learning. Typical approaches keep some exemplars for the past classes and use distillation regularization to retain the classification capability on the past classes and balance the past and new classes. However, there are four main problems with these approaches. First, the loss function is not efficient for classification. Second, there is unbalance problem between the past and new classes. Third, the size of pre-decided exemplars is usually limited and they might not be distinguishable from unseen new classes. Forth, the exemplars may not be allowed to be kept for a long time due to privacy regulations. To address these problems, we propose (a) a new loss function to combine the cross-entropy loss and distillation loss, (b) a simple way to estimate and remove the unbalance between the old and new classes , and (c) using Generative Adversarial Networks (GANs) to generate historical data and select representative exemplars during generation. We believe that the data generated by GANs have much less privacy issues than real images because GANs do not directly copy any real image patches. We evaluate the proposed method on CIFAR-100, Flower-102, and MS-Celeb-1M-Base datasets and extensive experiments demonstrate the effectiveness of our method.

cs.CV↗

arXiv2017

HP-GAN: Probabilistic 3D human motion prediction via GAN

Predicting and understanding human motion dynamics has many applications, such as motion synthesis, augmented reality, security, and autonomous vehicles. Due to the recent success of generative adversarial networks (GAN), there has been much interest in probabilistic estimation and synthetic data generation using deep neural network architectures and learning algorithms. We propose a novel sequence-to-sequence model for probabilistic human motion prediction, trained with a modified version of improved Wasserstein generative adversarial networks (WGAN-GP), in which we use a custom loss function designed for human motion prediction. Our model, which we call HP-GAN, learns a probability density function of future human poses conditioned on previous poses. It predicts multiple sequences of possible future human poses, each from the same input sequence but a different vector z drawn from a random distribution. Furthermore, to quantify the quality of the non-deterministic predictions, we simultaneously train a motion-quality-assessment model that learns the probability that a given skeleton sequence is a real human motion. We test our algorithm on two of the largest skeleton datasets: NTURGB-D and Human3.6M. We train our model on both single and multiple action types. Its predictive power for long-term motion estimation is demonstrated by generating multiple plausible futures of more than 30 frames from just 10 frames of input. We show that most sequences generated from the same input have more than 50\% probabilities of being judged as a real human sequence. We will release all the code used in this paper to Github.

cs.CV↗