arXiv subjects

Wendong Zhang

Publications and source records attributed to Wendong Zhang.

23 records · Page 2Linked to original sources

arXiv2021

Context-Aware Image Inpainting with Learned Semantic Priors

Recent advances in image inpainting have shown impressive results for generating plausible visual details on rather simple backgrounds. However, for complex scenes, it is still challenging to restore reasonable contents as the contextual information within the missing regions tends to be ambiguous. To tackle this problem, we introduce pretext tasks that are semantically meaningful to estimating the missing contents. In particular, we perform knowledge distillation on pretext models and adapt the features to image inpainting. The learned semantic priors ought to be partially invariant between the high-level pretext task and low-level image inpainting, which not only help to understand the global context but also provide structural guidance for the restoration of local textures. Based on the semantic priors, we further propose a context-aware image inpainting model, which adaptively integrates global semantics and local features in a unified image generator. The semantic learner and the image generator are trained in an end-to-end manner. We name the model SPL to highlight its ability to learn and leverage semantic priors. It achieves the state of the art on Places2, CelebA, and Paris StreetView datasets.

cs.CV↗

arXiv2019

Learning Context Graph for Person Search

Person re-identification has achieved great progress with deep convolutional neural networks. However, most previous methods focus on learning individual appearance feature embedding, and it is hard for the models to handle difficult situations with different illumination, large pose variance and occlusion. In this work, we take a step further and consider employing context information for person search. For a probe-gallery pair, we first propose a contextual instance expansion module, which employs a relative attention module to search and filter useful context information in the scene. We also build a graph learning framework to effectively employ context pairs to update target similarity. These two modules are built on top of a joint detection and instance feature learning framework, which improves the discriminativeness of the learned features. The proposed framework achieves state-of-the-art performance on two widely used person search datasets.

cs.CV↗

arXiv2017

Depth Structure Preserving Scene Image Generation

Key to automatically generate natural scene images is to properly arrange among various spatial elements, especially in the depth direction. To this end, we introduce a novel depth structure preserving scene image generation network (DSP-GAN), which favors a hierarchical and heterogeneous architecture, for the purpose of depth structure preserving scene generation. The main trunk of the proposed infrastructure is built on a Hawkes point process that models the spatial dependency between different depth layers. Within each layer generative adversarial sub-networks are trained collaboratively to generate realistic scene components, conditioned on the layer information produced by the point process. We experiment our model on a sub-set of SUNdataset with annotated scene images and demonstrate that our models are capable of generating depth-realistic natural scene image.

cs.CV↗

arXiv2016

Systematic theoretical analysis of dual-parameters RF readout by a novel LC-type passive sensor

This paper systematically studied the simultaneous measurement of two parameters by a LC-type passive sensor from the theoretical perspective. Based on the lumped circuit model of the typical LC-type passive dual-parameter sensor system, the influencing factors of the signal strength of the sensor as well as the influencing factors of signal crosstalk were both analyzed. It is found that the influencing factors of the RF readout signal strength of the sensor are mainly quality factors (Q factors) of the LC tanks, coupling coefficients, and the resonant frequency interval of the two LC tanks. And the influencing factors of the signal crosstalk are mainly coupling coefficient between the sensor inductance coils and the resonant frequency interval of the two LC tanks. The specific influence behavior of corresponding influencing factors on the signal strength and crosstalk is illustrated by a series of curves from numerical results simulated by using MATLAB software. Additionally, a decoupling scheme for solving the crosstalk problem algorithmically was proposed and a corresponding function was derived out. Overall, the theoretical analysis conducted in this work can provide design guidelines for making the dual-parameter LC-type passive sensor useful in practical applications.

physics.ins-det↗

arXiv2014

Physical white chaos generation

Physical chaos is a fascinating prospect for high-speed data security by serving as a masking carrier or a key source, but suffers from a colored spectrum that divulges system's intrinsic oscillations and degrades randomness. Here, we demonstrate that physical chaos with a white spectrum can be achieved by the optical heterodyning of two delayed-feedback lasers. A white chaotic spectrum with 1-dB fluctuation in a band of 11 GHz is experimentally obtained. The white chaos also has a perfect delta-like autocorrelation function and a high dimensionality of greater than 10, which makes chaos reconstruction extremely difficult and thus improves security.

physics.optics↗