arXiv subjects

Kun Zhao

Publications and source records attributed to Kun Zhao.

At least 73 records · Page 4Linked to original sources

arXiv2022

Few-Shot Class-Incremental Learning from an Open-Set Perspective

The continual appearance of new objects in the visual world poses considerable challenges for current deep learning methods in real-world deployments. The challenge of new task learning is often exacerbated by the scarcity of data for the new categories due to rarity or cost. Here we explore the important task of Few-Shot Class-Incremental Learning (FSCIL) and its extreme data scarcity condition of one-shot. An ideal FSCIL model needs to perform well on all classes, regardless of their presentation order or paucity of data. It also needs to be robust to open-set real-world conditions and be easily adapted to the new tasks that always arise in the field. In this paper, we first reevaluate the current task setting and propose a more comprehensive and practical setting for the FSCIL task. Then, inspired by the similarity of the goals for FSCIL and modern face recognition systems, we propose our method -- Augmented Angular Loss Incremental Classification or ALICE. In ALICE, instead of the commonly used cross-entropy loss, we propose to use the angular penalty loss to obtain well-clustered features. As the obtained features not only need to be compactly clustered but also diverse enough to maintain generalization for future incremental classes, we further discuss how class augmentation, data augmentation, and data balancing affect classification performance. Experiments on benchmark datasets, including CIFAR100, miniImageNet, and CUB200, demonstrate the improved performance of ALICE over the state-of-the-art FSCIL methods.

cs.CV↗

arXiv2022

Unsupervised Quantized Prosody Representation for Controllable Speech Synthesis

In this paper, we propose a novel prosody disentangle method for prosodic Text-to-Speech (TTS) model, which introduces the vector quantization (VQ) method to the auxiliary prosody encoder to obtain the decomposed prosody representations in an unsupervised manner. Rely on its advantages, the speaking styles, such as pitch, speaking velocity, local pitch variance, etc., are decomposed automatically into the latent quantize vectors. We also investigate the internal mechanism of VQ disentangle process by means of a latent variables counter and find that higher value dimensions usually represent prosody information. Experiments show that our model can control the speaking styles of synthesis results by directly manipulating the latent variables. The objective and subjective evaluations illustrated that our model outperforms the popular models.

eess.AS↗

arXiv2022

Context-Aware Scene Prediction Network (CASPNet)

Predicting the future motion of surrounding road users is a crucial and challenging task for autonomous driving (AD) and various advanced driver-assistance systems (ADAS). Planning a safe future trajectory heavily depends on understanding the traffic scene and anticipating its dynamics. The challenges do not only lie in understanding the complex driving scenarios but also the numerous possible interactions among road users and environments, which are practically not feasible for explicit modeling. In this work, we tackle the above challenges by jointly learning and predicting the motion of all road users in a scene, using a novel convolutional neural network (CNN) and recurrent neural network (RNN) based architecture. Moreover, by exploiting grid-based input and output data structures, the computational cost is independent of the number of road users and multi-modal predictions become inherent properties of our proposed method. Evaluation on the nuScenes dataset shows that our approach reaches state-of-the-art results in the prediction benchmark.

cs.CV↗

arXiv2021

The impact of data volume on performance of deep learning based building rooftop extraction using very high spatial resolution aerial images

Building rooftop data are of importance in several urban applications and in natural disaster management. In contrast to traditional surveying and mapping, by using high spatial resolution aerial images, deep learning-based building rooftops extraction methods are efficient and accurate. Although more training data is preferred in deep learning-based tasks, the effect of data volume on building extraction models is underexplored. Therefore, the paper explores the impact of data volume on the performance of building rooftop extraction from very-high-spatial-resolution (VHSR) images using deep learning-based methods. To do so, we manually labelled 0.12m spatial resolution aerial images and perform a comparative analysis of models trained on datasets of different sizes using popular deep learning architectures for segmentation tasks, including Fully Convolutional Networks (FCN)-8s, U-Net and DeepLabv3+. The experiments showed that with more training data, algorithms converged faster and achieved higher accuracy, while better algorithms were able to better mitigate the lack of training data.

cs.CV↗

arXiv2021

Maneuver-based Trajectory Prediction for Self-driving Cars Using Spatio-temporal Convolutional Networks

The ability to predict the future movements of other vehicles is a subconscious and effortless skill for humans and key to safe autonomous driving. Therefore, trajectory prediction for autonomous cars has gained a lot of attention in recent years. It is, however, still a hard task to achieve human-level performance. Interdependencies between vehicle behaviors and the multimodal nature of future intentions in a dynamic and complex driving environment render trajectory prediction a challenging problem. In this work, we propose a new, data-driven approach for predicting the motion of vehicles in a road environment. The model allows for inferring future intentions from the past interaction among vehicles in highway driving scenarios. Using our neighborhood-based data representation, the proposed system jointly exploits correlations in the spatial and temporal domain using convolutional neural networks. Our system considers multiple possible maneuver intentions and their corresponding motion and predicts the trajectory for five seconds into the future. We implemented our approach and evaluated it on two highway datasets taken in different countries and are able to achieve a competitive prediction performance.

cs.RO↗

arXiv2021

Eliminating the Barriers: Demystifying Wi-Fi Baseband Design and Introducing the PicoScenes Wi-Fi Sensing Platform

The research on Wi-Fi sensing has been thriving over the past decade but the process has not been smooth. Three barriers always hamper the research: unknown baseband design and its influence, inadequate hardware, and the lack of versatile and flexible measurement software. This paper tries to eliminate these barriers through the following work. First, we present an in-depth study of the baseband design of the Qualcomm Atheros AR9300 (QCA9300) NIC. We identify a missing item of the existing CSI model, namely, the CSI distortion, and identify the baseband filter as its origin. We also propose a distortion removal method. Second, we reintroduce both the QCA9300 and software-defined radio (SDR) as powerful hardware for research. For the QCA9300, we unlock the arbitrary tuning of both the carrier frequency and bandwidth. For SDR, we develop a high?performance software implementation of the 802.11a/g/n/ac/ax baseband, allowing users to fully control the baseband and access the complete physical-layer information. Third, we release the PicoScenes software, which supports concurrent CSI measure?ment from multiple QCA9300, Intel Wireless Link (IWL5300) and SDR hardware. PicoScenes features rich low-level controls, packet injection and software baseband implementation. It also allows users to develop their own measurement plugins. Finally, we report state-of-the-art results in the extensive evaluations of the PicoScenes system, such as the >2 GHz available spectrum on the QCA9300, concurrent CSI measurement, and up to 40 kHz and 1 kHz CSI measurement rates achieved by the QCA9300 and SDR. PicoScenes is available at https://ps.zpj.io.

cs.AR↗

arXiv2021

DIODE: Dilatable Incremental Object Detection

To accommodate rapid changes in the real world, the cognition system of humans is capable of continually learning concepts. On the contrary, conventional deep learning models lack this capability of preserving previously learned knowledge. When a neural network is fine-tuned to learn new tasks, its performance on previously trained tasks will significantly deteriorate. Many recent works on incremental object detection tackle this problem by introducing advanced regularization. Although these methods have shown promising results, the benefits are often short-lived after the first incremental step. Under multi-step incremental learning, the trade-off between old knowledge preserving and new task learning becomes progressively more severe. Thus, the performance of regularization-based incremental object detectors gradually decays for subsequent learning steps. In this paper, we aim to alleviate this performance decay on multi-step incremental detection tasks by proposing a dilatable incremental object detector (DIODE). For the task-shared parameters, our method adaptively penalizes the changes of important weights for previous tasks. At the same time, the structure of the model is dilated or expanded by a limited number of task-specific parameters to promote new task learning. Extensive experiments on PASCAL VOC and COCO datasets demonstrate substantial improvements over the state-of-the-art methods. Notably, compared with the state-of-the-art methods, our method achieves up to 6.0% performance improvement by increasing the number of parameters by just 1.2% for each newly learned task.

cs.CV↗

arXiv2021

Cascaded Residual Density Network for Crowd Counting

Crowd counting is a challenging task due to the issues such as scale variation and perspective variation in real crowd scenes. In this paper, we propose a novel Cascaded Residual Density Network (CRDNet) in a coarse-to-fine approach to generate the high-quality density map for crowd counting more accurately. (1) We estimate the residual density maps by multi-scale pyramidal features through cascaded residual density modules. It can improve the quality of density map layer by layer effectively. (2) A novel additional local count loss is presented to refine the accuracy of crowd counting, which reduces the errors of pixel-wise Euclidean loss by restricting the number of people in the local crowd areas. Experiments on two public benchmark datasets show that the proposed method achieves effective improvement compared with the state-of-the-art methods.

cs.CV↗

arXiv2021

On 3D Hall-MHD equations with fractional Laplacians: global well-posedness

Cauchy problem for 3D incompressible Hall-magnetohydrodynamics (Hall-MHD) system with fractional Laplacians is studied. First, global well-posedness of small-energy solutions with general initial data in $H^s$, $s>\frac{5}{2}$, is proved. Second, a special class of large-energy initial data is constructed, with which the Cauchy problem is globally well-posed. The proofs rely upon a new global bound of energy estimates involving Littlewood-Paley decomposition and Sobolev inequalities, which enables one to overcome the $\frac{1}{2}$-order derivative loss of the magnetic field.

math.AP↗

arXiv2021

On a Class of Nonlocal SIR Models

We revisit the classic Susceptible-Infected-Recovered (SIR) epidemic model and one of its nonlocal variations recently developed in \cite{Guan}. We introduce several new approaches to derive exact analytical solutions in the classical situation and analyze the corresponding effective approximations in the nonlocal setting. An interesting new feature of the nonlocal models, compared with the classic SIR model, is the appearance of multiple peak solutions for the infected population. We provide several rigorous results on the existence and non-existence of peak solutions with sharp asymptotics.

math.AP↗

arXiv2021

Polynomial Trajectory Predictions for Improved Learning Performance

The rising demand for Active Safety systems in automotive applications stresses the need for a reliable short to mid-term trajectory prediction. Anticipating the unfolding path of road users, one can act to increase the overall safety. In this work, we propose to train artificial neural networks for movement understanding by predicting trajectories in their natural form, as a function of time. Predicting polynomial coefficients allows us to increased accuracy and improve generalisation.

cs.CV↗

arXiv2021

Improving Conversational Recommendation System by Pretraining on Billions Scale of Knowledge Graph

Conversational Recommender Systems (CRSs) in E-commerce platforms aim to recommend items to users via multiple conversational interactions. Click-through rate (CTR) prediction models are commonly used for ranking candidate items. However, most CRSs are suffer from the problem of data scarcity and sparseness. To address this issue, we propose a novel knowledge-enhanced deep cross network (K-DCN), a two-step (pretrain and fine-tune) CTR prediction model to recommend items. We first construct a billion-scale conversation knowledge graph (CKG) from information about users, items and conversations, and then pretrain CKG by introducing knowledge graph embedding method and graph convolution network to encode semantic and structural information respectively.To make the CTR prediction model sensible of current state of users and the relationship between dialogues and items, we introduce user-state and dialogue-interaction representations based on pre-trained CKG and propose K-DCN.In K-DCN, we fuse the user-state representation, dialogue-interaction representation and other normal feature representations via deep cross network, which will give the rank of candidate items to be recommended.We experimentally prove that our proposal significantly outperforms baselines and show it's real application in Alime.

cs.IR↗

arXiv2021

Scalable Bayesian Deep Learning with Kernel Seed Networks

This paper addresses the scalability problem of Bayesian deep neural networks. The performance of deep neural networks is undermined by the fact that these algorithms have poorly calibrated measures of uncertainty. This restricts their application in high risk domains such as computer aided diagnosis and autonomous vehicle navigation. Bayesian Deep Learning (BDL) offers a promising method for representing uncertainty in neural network. However, BDL requires a separate set of parameters to store the mean and standard deviation of model weights to learn a distribution. This results in a prohibitive 2-fold increase in the number of model parameters. To address this problem we present a method for performing BDL, namely Kernel Seed Networks (KSN), which does not require a 2-fold increase in the number of parameters. KSNs use 1x1 Convolution operations to learn a compressed latent space representation of the parameter distribution. In this paper we show how this allows KSNs to outperform conventional BDL methods while reducing the number of required parameters by up to a factor of 6.6.

cs.LG↗

arXiv2021

Genetic Algorithm based hyper-parameters optimization for transfer Convolutional Neural Network

Hyperparameter optimization is a challenging problem in developing deep neural networks. Decision of transfer layers and trainable layers is a major task for design of the transfer convolutional neural networks (CNN). Conventional transfer CNN models are usually manually designed based on intuition. In this paper, a genetic algorithm is applied to select trainable layers of the transfer model. The filter criterion is constructed by accuracy and the counts of the trainable layers. The results show that the method is competent in this task. The system will converge with a precision of 97% in the classification of Cats and Dogs datasets, in no more than 15 generations. Moreover, backward inference according the results of the genetic algorithm shows that our method can capture the gradient features in network layers, which plays a part on understanding of the transfer AI models.

cs.NE↗

arXiv2020

SID: Incremental Learning for Anchor-Free Object Detection via Selective and Inter-Related Distillation

Incremental learning requires a model to continually learn new tasks from streaming data. However, traditional fine-tuning of a well-trained deep neural network on a new task will dramatically degrade performance on the old task -- a problem known as catastrophic forgetting. In this paper, we address this issue in the context of anchor-free object detection, which is a new trend in computer vision as it is simple, fast, and flexible. Simply adapting current incremental learning strategies fails on these anchor-free detectors due to lack of consideration of their specific model structures. To deal with the challenges of incremental learning on anchor-free object detectors, we propose a novel incremental learning paradigm called Selective and Inter-related Distillation (SID). In addition, a novel evaluation metric is proposed to better assess the performance of detectors under incremental learning conditions. By selective distilling at the proper locations and further transferring additional instance relation knowledge, our method demonstrates significant advantages on the benchmark datasets PASCAL VOC and COCO.

cs.CV↗

arXiv2020

Two-photon MINFLUX with doubled localization precision

Achieving localization with molecular precision has been of great interest for extending fluorescence microscopy to nanoscopy. MINFLUX pioneers this transition through point spread function (PSF) engineering, yet its performance is primarily limited by the signal-to-background ratio. Here we demonstrate that applying two-photon excitation to MINFLUX would double its localization precision through PSF engineering by nonlinear effect. Cramér-Rao Bound (CRB) is studied as the maximum localization precision, and CRB of two-photon MINFLUX is halved compared to single-photon MINFLUX in all three dimensions. Meanwhile, in order to achieve same localization precision with single-photon MINFLUX, two-photon MINFLUX requires only 1/4 of fluorescence photons, contributing to a possible 4-fold temporal resolution. Benefitted from two-photon excitation, registration-free multicolor nanoscopy and ultrafast color tracking can be achieved. Localization precision of MINFLUX can also be doubled using excitation with second-order Laguerre-Gaussian beams but would suffer from high fluorescence background compared to single-photon and two-photon first-order MINFLUX.

physics.optics↗

arXiv2020

Bounding Boxes Are All We Need: Street View Image Classification via Context Encoding of Detected Buildings

Street view images classification aiming at urban land use analysis is difficult because the class labels (e.g., commercial area), are concepts with higher abstract level compared to the ones of general visual tasks (e.g., persons and cars). Therefore, classification models using only visual features often fail to achieve satisfactory performance. In this paper, a novel approach based on a "Detector-Encoder-Classifier" framework is proposed. Instead of using visual features of the whole image directly as common image-level models based on convolutional neural networks (CNNs) do, the proposed framework firstly obtains the bounding boxes of buildings in street view images from a detector. Their contextual information such as the co-occurrence patterns of building classes and their layout are then encoded into metadata by the proposed algorithm "CODING" (Context encOding of Detected buildINGs). Finally, these bounding box metadata are classified by a recurrent neural network (RNN). In addition, we made a dual-labeled dataset named "BEAUTY" (Building dEtection And Urban funcTional-zone portraYing) of 19,070 street view images and 38,857 buildings based on the existing BIC GSV [1]. The dataset can be used not only for street view image classification, but also for multi-class building detection. Experiments on "BEAUTY" show that the proposed approach achieves a 12.65% performance improvement on macro-precision and 12% on macro-recall over image-level CNN based models. Our code and dataset are available at https://github.com/kyle-one/Context-Encoding-of-Detected-Buildings/

cs.CV↗

arXiv2020

Faster ILOD: Incremental Learning for Object Detectors based on Faster RCNN

The human vision and perception system is inherently incremental where new knowledge is continually learned over time whilst existing knowledge is retained. On the other hand, deep learning networks are ill-equipped for incremental learning. When a well-trained network is adapted to new categories, its performance on the old categories will dramatically degrade. To address this problem, incremental learning methods have been explored which preserve the old knowledge of deep learning models. However, the state-of-the-art incremental object detector employs an external fixed region proposal method that increases overall computation time and reduces accuracy comparing to Region Proposal Network (RPN) based object detectors such as Faster RCNN. The purpose of this paper is to design an efficient end-to-end incremental object detector using knowledge distillation. We first evaluate and analyze the performance of the RPN-based detector with classic distillation on incremental detection tasks. Then, we introduce multi-network adaptive distillation that properly retains knowledge from the old categories when fine-tuning the model for new task. Experiments on the benchmark datasets, PASCAL VOC and COCO, demonstrate that the proposed incremental detector based on Faster RCNN is more accurate as well as being 13 times faster than the baseline detector.

cs.CV↗