University of Nottingham Ningbo China
School of Computer Science, Ph.D. student
Computer Vision, Deep Learning
Supervisor: Guoping Qiu
2014 - 2015
University of Nottingham, Ph.D. student
Weiboyi, Beijing, Internship
Data Mining Group
China University of Mining & Technology Beijing
Master's Degree, Major in (Coal) Geology
Supervisor: Yuegang Tang
China University of Mining & Technology Beijing
Bachelor's Degree, Major in Geography
I am Xianxu Hou (侯贤旭), a PhD student at University of Nottingham working on Deep Learning and It’s application in Computer Vision.
Deep Feature Consistent Deep Image Transformations: Downscaling, Decolorization and HDR Tone Mapping
Building on crucial insights into the determining factors of the visual integrity of an image and the property of deep convolutional neural network (CNN), we have developed the Deep Feature Consistent Deep Image Transformation (DFC-DIT) framework which unifies challenging one-to-many mapping image processing problems such as image downscaling, decolorization (colour to grayscale conversion) and high dynamic range (HDR) image tone mapping. DFC-DIT uses one CNN as a non-linear mapper to transform an input image to an output image following the deep feature consistency principle which is enforced through another pretrained and fixed deep CNN. Unlike applications in related literature, none of these problems has a known ground truth or target for a supervised learning system. For each problem, we reason about a suitable learning objective function and develop an effective solution under the DFC-DIT framework. This is the first work that uses deep learning to solve and unify these three common image processing tasks. We present experimental results to demonstrate the effectiveness of the DFC-DIT technique and its state of the art performances.
Xianxu Hou, Jiang Duan, Guoping Qiu
Image Companding and Inverse Halftoning using Deep Convolutional Neural Networks
This paper presents a deep learning technology for tackling two traditional low-level image processing problems, companding and inverse halftoning. First, to the best knowledge of the authors, this is the first work that has successfully developed deep learning based solutions to these two traditional low-level image processing problems. Second, building on insights into the properties of visual quality of images and the internal representation properties of a deep convolutional neural network (CNN) and inspired by recent success of deep learning in other image processing applications, this paper has developed an effective deep learning method that trains a deep CNN as a nonlinear transformation function to map a lower bit depth image to higher bit depth or from a halftone image to a continuous tone image, and at the same time employs another pretrained deep CNN as a feature extractor to derive visually important features to construct the objective function for the training of the transformation CNN. Extensive experimental results are presented to show that the new deep learning based solution significantly outperforms previous methods and achieves new state-of-the-art results.
Xianxu Hou, Guoping Qiu
Learning Deep Semantic Attributes For User Video Summarization
This paper presents a Semantic Attribute assisted video summarization framework (SASUM). Compared with traditional methods, SASUM has several innovative features. Firstly, we use a natural language processing tool to discover a set of keywords from an image and text corpora to form the semantic attributes of visual contents. Secondly, we train a deep convolution neural network to extract visual features as well as predict the semantic attributes of video segments which enables us to represent video contents with visual and semantic features simultaneously. Thirdly, we construct a temporally constrained video segment affinity matrix and use a partially near duplicate image discovery technique to cluster visually and semantically consistent video frames together. These frame clusters can then be condensed to form an informative and compact summary of the video. We will present experimental results to show the effectiveness of the semantic attributes in assisting the visual features in video summarization and our new technique achieves state-of-the-art performance.
Ke Sun, Jiasong Zhu, Zhuo Lei, Xianxu Hou,Qian Zhang, Jiang Duan, Guoping Qiu
Deep Feature Consistent Variational Autoencoder
We present a novel method for constructing Variational Autoencoder (VAE). Instead of using pixel-by-pixel loss, we enforce deep feature consistency between the input and the output of a VAE, which ensures the VAE's output to preserve the spatial correlation characteristics of the input, thus leading the output to have a more natural visual appearance and better perceptual quality. Based on recent deep learning works such as style transfer, we employ a pre-trained deep convolutional neural network (CNN) and use its hidden features to define a feature perceptual loss for VAE training. Evaluated on the CelebA face dataset, we show that our model produces better results than other methods in the literature. We also show that our method can produce latent vectors that can capture the semantic information of face expressions and can be used to achieve state-of-the-art performance in facial attribute prediction.
Xianxu Hou, Linlin Shen, Ke Sun, Guoping Qiu
Object Specific Deep Feature and Its Application to Face Detection
We present a method for discovering and exploiting object specific deep features and use face detection as a case study. Motivated by the observation that certain convolutional channels of a Convolutional Neural Network (CNN) exhibit object specific responses, we seek to discover and exploit the convolutional channels of a CNN in which neurons are activated by the presence of specific objects in the input image. A method for explicitly fine-tuning a pre-trained CNN to induce an object specific channel (OSC) and systematically identifying it for the human face object has been developed. Building on the basic OSC features, we introduce a multi-scale approach to constructing robust face heatmaps for rapidly filtering out non-face regions thus significantly improving search efficiency for face detection in unconstrained settings. We show that multi-scale OSC can be used to develop simple and compact face detectors with state of the art performance.
Xianxu Hou, Ke Sun, Linlin Shen, Guoping Qiu
MVA 2017 (Oral)
Automatic Visual Theme Discovery from Joint Image and Text Corpora
We propose an unsupervised visual theme discovery framework as a better alternative for semantic representation of visual contents. We first show that tag based annotation lacks consistency and compactness for describing visually similar contents. We then learn the visual similarity between tags based on the visual features of the images containing the tags. At the same time, we use a natural language processing technique to measure the semantic similarity between tags. Finally, we cluster tags into visual themes based on their visual similarity and semantic similarity measures using a spectral clustering algorithm. We then design three common computer vision tasks, example based image search, keyword based image search and image labeling to explore potential application of our visual themes discovery framework. In experiments, visual themes significantly outperform tags on semantic image understanding and achieve state-of-art performance in all three tasks.
Ke Sun, Xianxu Hou, Qian Zhang, Guoping Qiu
ICMIP 2017 (Oral)
Some publications in (Coal) Geology
Geochemistry of Tuffs and Tonstein from the Mayi Exploration Area, Guizhou China
Mineralogical and geochemical characteristics of the Late Permian tuffs and tonsteins are studied by X-ray fluorescence(XRF),powder X-ray diffraction(XRD),and inductively-coupled plasma mas spectrometry (ICP-MS).The results show that clay minerals in most tuffs and tonsteins are dominated by kaolinite. All the samples are rich in trace elements Nb, Ta, Zr, Hf and REEs, and contains high content of minerals like pyrite due to the seawater influences. The primitive-mantle normalized spider-diagrams are similar to the ocean island basalt (OIB), which indicates that the volcanic ash forming tuffs and tonsteins may originate from a waning mantle plume, furthermore the ratio of Nb/Ta suggests a contamination from continental crust or lithosphere.
Xianxu Hou, Yuegang Tang, Cong Wang, Weicheng Gao, Qiang Wei
Bulletin of Mineralogy, Petrology and Geochemistry, 2014
Coal Petrology and Coal facies of Zhongliangshan Mining Area, Chongqing China
The primary coal seams in Zhongliangshan mining area are analyzed by macerals ternary diagrams and coal facies diagrams based on the maceral classification of ICCP 1994 system. The coal seams are characterized by the predominance of semi-bright and structural coals, which is a result of Epigenetic tectonic activity. The petrographic composition is dominated by vitrinite, and collodetrinite and vitrodetrinite are the major macerals. The sedimentary environments is deduced from petrographic composition and coal facies diagrams demonstrate the transition from low delta plain to limnic environment, and the mires were influenced by marine flooding.
Xianxu Hou, Yuegang Tang, Xiaoxia Song, Mingxian Yang, Mingtao Guo, Long Jia
Coal Geology & Exploration, 2013
Taiyuan and Shanxi Formations Coal Rank Distribution and Metamorphism Analysis in Shanxi Province
Coals in The Shanxi Province China have complete coal ranks and excellent quality and the province is the main coking coal and anthracite base in the country. Based on coal resource prediction data and mining areas exploration data in many years, this paper summarizes the coal rank distribution and metamorphism features in Taiyuan and Shanxi Formations.
Xichao Xie, Qinghui Zhang, Yuegang Tang, Zhengxi Zhang, Haisheng Wang, Xianxu Hou, Xiaodong Chen, Yufei Su, Cong Wang, Long Jia
Coal Geology of China, 2011