arXiv subjects

Huan Hu

Publications and source records attributed to Huan Hu.

28 records · Page 2Linked to original sources

arXiv2022

Data-Centric AI Paradigm Based on Application-Driven Fine-Grained Dataset Design

Deep learning has a wide range of applications in industrial scenario, but reducing false alarm (FA) remains a major difficulty. Optimizing network architecture or network parameters is used to tackle this challenge in academic circles, while ignoring the essential characteristics of data in application scenarios, which often results in increased FA in new scenarios. In this paper, we propose a novel paradigm for fine-grained design of datasets, driven by industrial applications. We flexibly select positive and negative sample sets according to the essential features of the data and application requirements, and add the remaining samples to the training set as uncertainty classes. We collect more than 10,000 mask-wearing recognition samples covering various application scenarios as our experimental data. Compared with the traditional data design methods, our method achieves better results and effectively reduces FA. We make all contributions available to the research community for broader use. The contributions will be available at https://github.com/huh30/OpenDatasets.

cs.CV↗

arXiv2021

Sublinear Time Nearest Neighbor Search over Generalized Weighted Manhattan Distance

Nearest Neighbor Search (NNS) over generalized weighted distances is fundamental to a wide range of applications. The problem of NNS over the generalized weighted square Euclidean distance has been studied in previous work. However, numerous studies have shown that the Manhattan distance could be more effective than the Euclidean distance for high-dimensional NNS, which indicates that the generalized weighted Manhattan distance is possibly more practical than the generalized weighted square Euclidean distance in high dimensions. To the best of our knowledge, no prior work solves the problem of NNS over the generalized weighted Manhattan distance in sublinear time. This paper achieves the goal by proposing two novel hashing schemes ($d_w^{l_1},l_2$)-ALSH and ($d_w^{l_1},θ$)-ALSH.

cs.DB↗

arXiv2020

Efficient Approximate Nearest Neighbor Search for Multiple Weighted $l_{p\leq2}$ Distance Functions

Nearest neighbor search is fundamental to a wide range of applications. Since the exact nearest neighbor search suffers from the "curse of dimensionality", approximate approaches, such as Locality-Sensitive Hashing (LSH), are widely used to trade a little query accuracy for a much higher query efficiency. In many scenarios, it is necessary to perform nearest neighbor search under multiple weighted distance functions in high-dimensional spaces. This paper considers the important problem of supporting efficient approximate nearest neighbor search for multiple weighted distance functions in high-dimensional spaces. To the best of our knowledge, prior work can only solve the problem for the $l_2$ distance. However, numerous studies have shown that the $l_p$ distance with $p\in(0,2)$ could be more effective than the $l_2$ distance in high-dimensional spaces. We propose a novel method, WLSH, to address the problem for the $l_p$ distance for $p\in(0,2]$. WLSH takes the LSH approach and can theoretically guarantee both the efficiency of processing queries and the accuracy of query results while minimizing the required total number of hash tables. We conduct extensive experiments on synthetic and real data sets, and the results show that WLSH achieves high performance in terms of query efficiency, query accuracy and space consumption.

cs.DB↗

arXiv2019

Feature Aggregation Network for Video Face Recognition

This paper aims to learn a compact representation of a video for video face recognition task. We make the following contributions: first, we propose a meta attention-based aggregation scheme which adaptively and fine-grained weighs the feature along each feature dimension among all frames to form a compact and discriminative representation. It makes the best to exploit the valuable or discriminative part of each frame to promote the performance of face recognition, without discarding or despising low quality frames as usual methods do. Second, we build a feature aggregation network comprised of a feature embedding module and a feature aggregation module. The embedding module is a convolutional neural network used to extract a feature vector from a face image, while the aggregation module consists of cascaded two meta attention blocks which adaptively aggregate the feature vectors into a single fixed-length representation. The network can deal with arbitrary number of frames, and is insensitive to frame order. Third, we validate the performance of proposed aggregation scheme. Experiments on publicly available datasets, such as YouTube face dataset and IJB-A dataset, show the effectiveness of our method, and it achieves competitive performances on both the verification and identification protocols.

cs.CV↗

arXiv2019

Video synthesis of human upper body with realistic face

This paper presents a generative adversarial learning-based human upper body video synthesis approach to generate an upper body video of target person that is consistent with the body motion, face expression, and pose of the person in source video. We use upper body keypoints, facial action units and poses as intermediate representations between source video and target video. Instead of directly transferring the source video to the target video, we firstly map the source person's facial action units and poses into the target person's facial landmarks, then combine the normalized upper body keypoints and generated facial landmarks with spatio-temporal smoothing to generate the corresponding target video's image. Experimental results demonstrated the effectiveness of our method.

cs.CV↗

arXiv2019

A Neural Virtual Anchor Synthesizer based on Seq2Seq and GAN Models

This paper presents a novel framework to generate realistic face video of an anchor, who is reading certain news. This task is also known as Virtual Anchor. Given some paragraphs of words, we first utilize a pretrained Word2Vec model to embed each word into a vector; then we utilize a Seq2Seq-based model to translate these word embeddings into action units and head poses of the target anchor; these action units and head poses will be concatenated with facial landmarks as well as the former $n$ synthesized frames, and the concatenation serves as input of a Pix2PixHD-based model to synthesize realistic facial images for the virtual anchor. The experimental results demonstrate our framework is feasible for the synthesis of virtual anchor.

cs.CV↗

arXiv2019

A Realistic Face-to-Face Conversation System based on Deep Neural Networks

To improve the experiences of face-to-face conversation with avatar, this paper presents a novel conversation system. It is composed of two sequence-to-sequence models respectively for listening and speaking and a Generative Adversarial Network (GAN) based realistic avatar synthesizer. The models exploit the facial action and head pose to learn natural human reactions. Based on the models' output, the synthesizer uses the Pixel2Pixel model to generate realistic facial images. To show the improvement of our system, we use a 3D model based avatar driving scheme as a reference. We train and evaluate our neural networks with the data from ESPN shows. Experimental results show that our conversation system can generate natural facial reactions and realistic facial images.

cs.CV↗

arXiv2017

Thermal graphene metamaterials and epsilon-near-zero high temperature plasmonics

The key feature of a thermophotovoltaic (TPV) emitter is the enhancement of thermal emission corresponding to energies just above the bandgap of the absorbing photovoltaic cell and simultaneous suppression of thermal emission below the bandgap. We show here that a single layer plasmonic coating can perform this task with high efficiency. Our key design principle involves tuning the epsilon-near-zero frequency (plasma frequency) of the metal acting as a thermal emitter to the electronic bandgap of the semiconducting cell. This approach utilizes the change in reflectivity of a metal near its plasma frequency (epsilon-near-zero frequency) to lead to spectrally selective thermal emission and can be adapted to large area coatings using high temperature plasmonic materials. We provide a detailed analysis of the spectral and angular performance of high temperature plasmonic coatings as TPV emitters. We show the potential of such high temperature plasmonic thermal emitter coatings (p-TECs) for narrowband near-field thermal emission. We also show the enhancement of near-surface energy density in graphene-multilayer thermal metamaterials due to a topological transition at an effective epsilon-near-zero frequency. This opens up spectrally selective thermal emission from graphene multilayers in the infrared frequency regime. Our design paves the way for the development of single layer p-TECs and graphene multilayers for spectrally selective radiative heat transfer applications.

physics.optics↗

arXiv2015

Optical Characterization of Epsilon-Near-Zero, Epsilon-Near-Pole and Hyperbolic Response in Nanowire Metamaterials

We report on the optical and physical characterization of metallic nanowire (NW) metamaterials fabricated by electrodeposition of ~30 nm diameter gold nanowires in nano-porous anodic aluminum oxide. We observe a uniaxial anisotropic dielectric response for the NW metamaterials that displays both epsilon-near-zero (ENZ) and epsilon-near-pole (ENP) resonances. We show that a fundamental difference in the behavior of NW-metamaterials from metal-dielectric multilayer (ML) metamaterials is the differing directions of the epsilon-near-zero (ENZ) and epsilon-near-pole (ENP) dielectric responses relative to the optical axis of the effective dielectric tensor. In contrast to multilayer metamaterials, nanowire metamaterials exhibit an omnidirectional ENP and an angularly dependent ENZ. Also in contrast to ML metamaterials, the NW metamaterials exhibit ENP and ENZ resonances that are highly absorptive and can be effectively excited from free space. Our fabrication allows a large tunability of the epsilon-near-zero resonance in the visible and near IR spectrum from 583 nm to 805 nm as the gold nanorod fill fraction changes from 26% to 10.5%. We support our fabrication process flow at each step with rigorous physical and optical characterization. Energy dispersive X-ray (EDX) and X-ray diffraction (XRD) analyses are used to ascertain the quality of electrochemically deposited Au nanowires prior to and after annealing. Our experimental results are in agreement with simulations of the periodic plasmonic crystal and also analytical calculations in the effective medium metamaterial limit. We also experimentally characterize the role of spatial dispersion at the ENZ resonance and show that the effect does not occur for the ENP resonance. The application of these materials to the fields of biosensing, quantum optics and thermal devices shows considerable promise.

cond-mat.mtrl-sci↗

arXiv2014

Thermal excitation of plasmons for near-field thermophotovoltaics

The traditional approaches of exciting plasmons consist of using electrons (eg: electron energy loss spectroscopy) or light (Kretchman and Otto geometry) while more recently plasmons have been excited even by single photons. A different approach: thermal excitation of a plasmon resonance at high temperatures using alternate plasmonic media was proposed by S. Molesky et.al., Opt. Exp. 21.101, A96-A110, (2013). Here, we show how the long-standing search for a high temperature narrow band near-field emitter for thermophotovoltaics can be fulfilled by high temperature plasmonics. We also describe how to control Wein's displacement law in the near-field using high temperature epsilon-near-zero metamaterials. Finally, we show that our work opens up an interesting direction of research for the field of slow light: thermal emission control.

physics.optics↗