Search

Hyperbolic geometry, a Riemannian manifold endowed with constant sectional negative curvature, has been considered an alternative embedding space in many learning scenarios, \eg, natural language processing, graph learning, \etc, as a result of its intriguing property of encoding the data's hierarchical structure (like irregular graph or tree-likeness data). Recent studies prove that such data hierarchy also exists in the visual dataset, and investigate the successful practice of hyperbolic geometry in the computer vision (CV) regime, ranging from the classical image classification to advanced model adaptation learning. This paper presents the first and most up-to-date literature review of hyperbolic spaces for CV applications. To this end, we first introduce the background of hyperbolic geometry, followed by a comprehensive investigation of algorithms, with geometric prior of hyperbolic space, in the context of visual applications. We also conclude this manuscript and identify possible future directions., First survey paper for the hyperbolic geometry in CV applications

Scene text image super-resolution (STISR) aims to simultaneously increase the resolution and legibility of the text images, and the resulting images will significantly affect the performance of downstream tasks. Although numerous progress has been made, existing approaches raise two crucial issues: (1) They neglect the global structure of the text, which bounds the semantic determinism of the scene text. (2) The priors, e.g., text prior or stroke prior, employed in existing works, are extracted from pre-trained text recognizers. That said, such priors suffer from the domain gap including low resolution and blurriness caused by poor imaging conditions, leading to incorrect guidance. Our work addresses these gaps and proposes a plug-and-play module dubbed Dual Prior Modulation Network (DPMN), which leverages dual image-level priors to bring performance gain over existing approaches. Specifically, two types of prior-guided refinement modules, each using the text mask or graphic recognition result of the low-quality SR image from the preceding layer, are designed to improve the structural clarity and semantic accuracy of the text, respectively. The following attention mechanism hence modulates two quality-enhanced images to attain a superior SR result. Extensive experiments validate that our method improves the image quality and boosts the performance of downstream tasks over five typical approaches on the benchmark. Substantial visualizations and ablation studies demonstrate the advantages of the proposed DPMN. Code is available at: https://github.com/jdfxzzy/DPMN., Comment: Accepted by AAAI-2023

This paper studies the problem of measuring and predicting how memorable an image is to pattern recognition machines, as a path to explore machine intelligence. Firstly, we propose a self-supervised machine memory quantification pipeline, dubbed ``MachineMem measurer'', to collect machine memorability scores of images. Similar to humans, machines also tend to memorize certain kinds of images, whereas the types of images that machines and humans memorize are different. Through in-depth analysis and comprehensive visualizations, we gradually unveil that``complex" images are usually more memorable to machines. We further conduct extensive experiments across 11 different machines (from linear classifiers to modern ViTs) and 9 pre-training methods to analyze and understand machine memory. This work proposes the concept of machine memorability and opens a new research direction at the interface between machine memory and visual data., Code: https://github.com/JunlinHan/MachineMem Project page: https://junlinhan.github.io/projects/machinemem.html

Learning a latent embedding to understand the underlying nature of data distribution is often formulated in Euclidean spaces with zero curvature. However, the success of the geometry constraints, posed in the embedding space, indicates that curved spaces might encode more structural information, leading to better discriminative power and hence richer representations. In this work, we investigate benefits of the curved space for analyzing anomalies or out-of-distribution objects in data. This is achieved by considering embeddings via three geometry constraints, namely, spherical geometry (with positive curvature), hyperbolic geometry (with negative curvature) or mixed geometry (with both positive and negative curvatures). Three geometric constraints can be chosen interchangeably in a unified design given the task at hand. Tailored for the embeddings in the curved space, we also formulate functions to compute the anomaly score. Two types of geometric modules (i.e., Geometric-in-One and Geometric-in-Two models) are proposed to plug in the original Euclidean classifier, and anomaly scores are computed from the curved embeddings. We evaluate the resulting designs under a diverse set of visual recognition scenarios, including image detection (multi-class OOD detection and one-class anomaly detection) and segmentation (multi-class anomaly segmentation and one-class anomaly segmentation). The empirical results show the effectiveness of our proposal through the consistent improvement over various scenarios.

Automated real-time evaluation of counselor-client interaction is important for ensuring quality counseling but the rules are difficult to articulate. Recent advancements in machine learning methods show the possibility of learning such rules automatically. However, these methods often demand large scale and high quality counseling data, which are difficult to collect. To address this issue, we build an online counseling platform, which allows professional psychotherapists to provide free counseling services to those are in need. In exchange, we collect the counseling transcripts. Within a year of its operation, we manage to get one of the largest set of (675) transcripts of counseling sessions. To further leverage the valuable data we have, we label our dataset using both coarse- and fine-grained labels and use a set of pretraining techniques. In the end, we are able to achieve practically useful accuracy in both labeling system., Comment: 15 pages, 4 figures

Current NLP techniques have been greatly applied in different domains. In this paper, we propose a human-in-the-loop framework for robotic grasping in cluttered scenes, investigating a language interface to the grasping process, which allows the user to intervene by natural language commands. This framework is constructed on a state-of-the-art rasping baseline, where we substitute a scene-graph representation with a text representation of the scene using BERT. Experiments on both simulation and physical robot show that the proposed method outperforms conventional object-agnostic and scene-graph based methods in the literature. In addition, we find that with human intervention, performance can be significantly improved., Comment: 15 pages, 10 figures, Coling2022 Oral

This work studies the joint rain and haze removal problem. In real-life scenarios, rain and haze, two often co-occurring common weather phenomena, can greatly degrade the clarity and quality of the scene images, leading to a performance drop in the visual applications, such as autonomous driving. However, jointly removing the rain and haze in scene images is ill-posed and challenging, where the existence of haze and rain and the change of atmosphere light, can both degrade the scene information. Current methods focus on the contamination removal part, thus ignoring the restoration of the scene information affected by the change of atmospheric light. We propose a novel deep neural network, named Asymmetric Dual-decoder U-Net (ADU-Net), to address the aforementioned challenge. The ADU-Net produces both the contamination residual and the scene residual to efficiently remove the rain and haze while preserving the fidelity of the scene information. Extensive experiments show our work outperforms the existing state-of-the-art methods by a considerable margin in both synthetic data and real-world data benchmarks, including RainCityscapes, BID Rain, and SPA-Data. For instance, we improve the state-of-the-art PSNR value by 2.26/4.57 on the RainCityscapes/SPA-Data, respectively. Codes will be made available freely to the research community., Comment: 12 pages, 35 figures

In this paper, we present and study a new image segmentation task, called Generalized Open-set Semantic Segmentation (GOSS). Previously, with the well-known open-set semantic segmentation (OSS), the intelligent agent only detects the unknown regions without further processing, limiting their perception of the environment. It stands to reason that a further analysis of the detected unknown pixels would be beneficial. Therefore, we propose GOSS, which unifies the abilities of two well-defined segmentation tasks, OSS and generic segmentation (GS), in a holistic way. Specifically, GOSS classifies pixels as belonging to known classes, and clusters (or groups) of pixels of unknown class are labelled as such. To evaluate this new expanded task, we further propose a metric which balances the pixel classification and clustering aspects. Moreover, we build benchmark tests on top of existing datasets and propose a simple neural architecture as a baseline, which jointly predicts pixel classification and clustering under open-set settings. Our experiments on multiple benchmarks demonstrate the effectiveness of our baseline. We believe our new GOSS task can produce an expressive image understanding for future research. Code will be made available.

Learning and generalizing from limited examples, i,e, few-shot learning, is of core importance to many real-world vision applications. A principal way of achieving few-shot learning is to realize an embedding where samples from different classes are distinctive. Recent studies suggest that embedding via hyperbolic geometry enjoys low distortion for hierarchical and structured data, making it suitable for few-shot learning. In this paper, we propose to learn a context-aware hyperbolic metric to characterize the distance between a point and a set associated with a learned set to set distance. To this end, we formulate the metric as a weighted sum on the tangent bundle of the hyperbolic space and develop a mechanism to obtain the weights adaptively and based on the constellation of the points. This not only makes the metric local but also dependent on the task in hand, meaning that the metric will adapt depending on the samples that it compares. We empirically show that such metric yields robustness in the presence of outliers and achieves a tangible improvement over baseline models. This includes the state-of-the-art results on five popular few-shot classification benchmarks, namely mini-ImageNet, tiered-ImageNet, Caltech-UCSD Birds-200-2011 (CUB), CIFAR-FS, and FC100., Comment: Accepted at AAAI2022

We propose and study a novel task named Blind Image Decomposition (BID), which requires separating a superimposed image into constituent underlying images in a blind setting, that is, both the source components involved in mixing as well as the mixing mechanism are unknown. For example, rain may consist of multiple components, such as rain streaks, raindrops, snow, and haze. Rainy images can be treated as an arbitrary combination of these components, some of them or all of them. How to decompose superimposed images, like rainy images, into distinct source components is a crucial step toward real-world vision systems. To facilitate research on this new task, we construct multiple benchmark datasets, including mixed image decomposition across multiple domains, real-scenario deraining, and joint shadow/reflection/watermark removal. Moreover, we propose a simple yet general Blind Image Decomposition Network (BIDeN) to serve as a strong baseline for future work. Experimental results demonstrate the tenability of our benchmarks and the effectiveness of BIDeN., ECCV 2022. Project page: https://junlinhan.github.io/projects/BID.html. Code: https://github.com/JunlinHan/BID

Signal modulation recognition is a key technique in the domain of spectrum monitoring and management. However, due to the complexity of signal modulation types and the lack of signal prior information, the performance of modulation recognition might degrade dramatically. In this paper, a modulation recognition approach based on enhanced data representation and neural network is proposed. Firstly, a deep convolutional neural network structure is designed for learning the latent features and identifies the modulation type of input signal. Then, in order to alleviate the information scarcity of received signal, the enhanced representation matrix of received spectrum data is constructed as networks inputs. Finally, the convolutional neural network is adjusted to adopt and combined with the enhanced data representation to form the presented modulation recognition scheme. The experiments results demonstrate that the proposed neural network is effective for modulation recognition and the enhanced data representation can effectively improve the performance of the neural network.

Sample-and-rank is a key decoding strategy for modern generation-based dialogue systems. It helps achieve diverse and high-quality responses by selecting an answer from a small pool of generated candidates. The current state-of-the-art ranking methods mainly use an encoding paradigm called Cross-Encoder, which separately encodes each context-candidate pair and ranks the candidates according to their fitness scores. However, Cross-Encoder repeatedly encodes the same lengthy context for each candidate, resulting in high computational costs. Poly-Encoder addresses the above problems by reducing the interaction between context and candidates, but with a price of performance drop. In this work, we develop a new paradigm called Uni-Encoder, that keeps the full attention over each pair as in Cross-Encoder while only encoding the context once, as in Poly-Encoder. Uni-Encoder encodes all the candidates with the context in one forward pass. We use the same positional embedding for all candidates to ensure they are treated equally and design a new attention mechanism to avoid confusion. Our Uni-Encoder can simulate other ranking paradigms using different attention and response concatenation methods. Extensive experiments show that our proposed paradigm achieves new state-of-the-art results on four benchmark datasets with high computational efficiency. For instance, it improves R10@1 by 2.9% with an approximately 4X faster inference speed on the Ubuntu V2 dataset., Accepted to the Findings of ACL 2023

Aiming at the problems of large amount of data, strong correlation between adjacent pixels and high redundancy in the process of color image encryption, a color image encryption algorithm based on hyperchaotic system and DNA coding operation is proposed, which combines with the neural network to generate a better fusion random sequence as key stream, and uses DNA transformation rules and block encryption mechanism to realize image scrambling diffusion encryption in pixel level. The simulation results show that the encryption algorithm has high security integrity and strong anti-attack ability.

Aiming at the problems of inaccurate judgment of linear range and the inability to measure dynamically of traditional compliance measuring instrument of loudspeaker spider，a spider compliance measurement system based on STM32 was designed. In the hardware design, the function of auto-measurement was realized by motion control circuit and data acquisition circuit, and a low voltage ripple DC voltage regulator was used to make the sampling result more accurate. In the part of the software design, the sliding mean filter algorithm was used to reduce the output error, the least square algorithm was used to improve the fitting goodness, and the linear approximation was used to obtain the linear range of the spider. The experiments show that the resolution of the system is high, the compliance curve and the linear range of the spider can be obtained accurately.

In order to explore bearing characteristics of a new-type static drill rooted nodular pile foundation (SDRN), which is composed of PHC pile, bamboo joint pile, and cement soils, field tests of three piles were carried out by embedding internal rebar stress gauges to collect test data. The test results show that the SDRN piles were under elastic state and the load-settlement curves changed slowly before reaching the ultimate capacity. As pile head loads increased, the pile shaft frictions developed progressively, and axial forces gradually reduced along pile depths. Taking into account the interaction between the pile, the cement soils, and surrounding soils, a simplified method for calculating settlements and bearing capacity of SDRN piles was proposed. With corresponding parameters, the computing results obtained by the proposed method were compared with the field experimental data, which indicates acceptable agreements; thus, it is concluded that the applicability and predictive capability of the proposed method were verified.

Learning and generalizing to novel concepts with few samples (Few-Shot Learning) is still an essential challenge to real-world applications. A principle way of achieving few-shot learning is to realize a model that can rapidly adapt to the context of a given task. Dynamic networks have been shown capable of learning content-adaptive parameters efficiently, making them suitable for few-shot learning. In this paper, we propose to learn the dynamic kernels of a convolution network as a function of the task at hand, enabling faster generalization. To this end, we obtain our dynamic kernels based on the entire task and each sample and develop a mechanism further conditioning on each individual channel and position independently. This results in dynamic kernels that simultaneously attend to the global information whilst also considering minuscule details available. We empirically show that our model improves performance on few-shot classification and detection tasks, achieving a tangible improvement over several baseline models. This includes state-of-the-art results on 4 few-shot classification benchmarks: mini-ImageNet, tiered-ImageNet, CUB and FC100 and competitive results on a few-shot detection dataset: MS COCO-PASCAL-VOC., Comment: ECCV2022

Learning and generalizing from limited examples, i,e, few-shot learning, is of core importance to many real-world vision applications. A principal way of achieving few-shot learning is to realize an embedding where samples from different classes are distinctive. Recent studies suggest that embedding via hyperbolic geometry enjoys low distortion for hierarchical and structured data, making it suitable for few-shot learning. In this paper, we propose to learn a context-aware hyperbolic metric to characterize the distance between a point and a set associated with a learned set to set distance. To this end, we formulate the metric as a weighted sum on the tangent bundle of the hyperbolic space and develop a mechanism to obtain the weights adaptively and based on the constellation of the points. This not only makes the metric local but also dependent on the task in hand, meaning that the metric will adapt depending on the samples that it compares. We empirically show that such metric yields robustness in the presence of outliers and achieves a tangible improvement over baseline models. This includes the state-of-the-art results on five popular few-shot classification benchmarks, namely mini-ImageNet, tiered-ImageNet, Caltech-UCSD Birds-200-2011 (CUB), CIFAR-FS, and FC100., Accepted at AAAI2022

Prior to the introduction of Graph Neural Networks (GNNs), modeling and analyzing irregular data, particularly graphs, was thought to be the Achilles' heel of deep learning. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbors. The core concept of GNNs is to find a representation by recursively aggregating the representations of a central node and those of its neighbor, and its success has been demonstrated by many GNNs' designs. However, most of them only focus on using the first-order information between a node and its neighbors. In this paper, we introduce a central node permutation variant function through a frustratingly simple and innocent-looking modification to the core operation of a GNN, namely the Feature cOrrelation aGgregation (FOG) module which learns the second-order information from feature correlation between a node and its neighbors in the pipeline. By adding FOG into existing variants of GNNs, we empirically verify this second-order information complements the features generated by original GNNs across a broad set of benchmarks. A tangible boost in performance of the model is observed where the model surpasses previous state-of-the-art results by a significant margin while employing fewer parameters. (e.g., 33.116% improvement on a real-world molecular dataset using graph convolutional networks).

Vaccines against Porcine circovirus type 2 (PCV2) have been studied intensely and found to be effective in decreasing mortality and improving growth in swine populations. In this study, interleukin-23 (IL-23) gene was cloned from peripheral blood mononuclear cells (PBMCs) of Tibetan pigs and inserted into a eukaryotic VR1020 expression vector-VRIL23. Coated with chitosan (CS), the VRIL23-CS was intramuscularly injected into 3-week-old piglets with PCV2 vaccine. The blood was collected after vaccination at 0, 1, 2, 4, 8, and 12 weeks, respectively, to detect the immunological changes. The IgG2a and specific PCV2 antibodies were detected using ELISA, and blood CD4+ and CD8+ T cells were quantified by flow cytometry. Quantitative fluorescence PCR was used to evaluate the expression of immune genes. The results indicate that leukocytes, erythrocytes, and CD4+ and CD8+ T cells increased significantly in the blood of VRIL23-CS inoculated piglets in comparison with the control (p &lt, 0.05) and so did the IgG2a and PCV2 antibodies. In addition, the expressions of Toll-like receptor (TLR) 2, TLR7, cluster of differentiation (CD) 45, IL-15, IL-12, signal transducer and activator of transcription (STAT)1, STAT2, STAT3, STAT4, and B-cell lymphoma (Bcl)-2 genes were also obviously higher in the VRIL23-CS inoculated pigs at different time points (p &lt, 0.05). Overall, the results demonstrated that VRIL23-CS can enhance the comprehensive immune responses to PCV2 vaccine in vivo and has the promising potential to be developed into a safe and effective adjuvant to promote the immunity of pig against PCV disease.

A 3D numerical heat transfer model is established and validated to investigate the effect of groundwater flow on the thermal response of geothermal energy pile (GEP) groups. A constant heating flux with different groundwater flow velocities is applied to four common types of GEP groups in residential buildings. The results show that groundwater flow has an effect on the thermal performance of GEP groups, and the average coefficient of performance of the GEP groups has an increase with groundwater flow. In the short-term (e.g., a month), the GEP spacing of 3 times the pile diameter is convenient from the viewpoint of energy when the flow velocity is less than 0.01 m/yr. The amount of heat is carried from upstream GEPs to the downstream by groundwater flow, and the temperature difference between the upstream and downstream GEPs becomes more obvious with the increase of groundwater flow velocity. In addition, we discussed the differential thermal expansion caused by groundwater flow and its effect on the mechanical behavior of GEPs.

A diode laser sensor based on absorption spectroscopy is developed for measurement of self-broadening coefficients for the three lines of v3 band of CO at high temperature. the spectra of CO at 6390.8152 cm-1, 6393.1767 cm-1 and 6395.4312 cm-1 at temperatures ranging from 323K to 873K with different pressures are recorded. The line widths are acquired by the line shape fittings based on Voigt function, and the collitional broadening coefficients and temperature dependences of the three lines are also derived.

In order to develop a novel effective immunomodulator to enhance pig resistance against post-weaning multi systemic wasting syndrome (PMWS), a recombinant plasmid co-expressing pig interleukin-2 (IL-2) and fusion interleukin-4/6 (IL-4/6) genes, designated VRIL4/6-2, was constructed and encapsulated in chitosan (CS) nanoparticles prepared by the ionotropic gelation method. Then 21-day old piglets were divided into two groups and intramuscularly injected respectively with VPIL4/6-2-CS and saline along with the porcine circovirus-2 (PCV-2) vaccine. The blood was collected from each piglet on days 0, 7, 14, 28, 56 and 84 after vaccination to assay the immunological changes. Content of IgG2a, CD4+, CD8+ T cells increased significantly in the sera or blood of piglets treated with VPIL4/6-2-CS (P

The polystyrene/rectorite nanocomposites were prepared by free radical polymerization of styrene containing dispersed organophilic rectorite. The structures and thermal properties of these hybrids have been investigated by X-ray diffraction (XRD), fourier transform infrared (FT-IR), positron annihilation spectroscopy (PAS) and thermal gravimetric analysis (TGA) techniques. It was found that exfoliation of rectorite in polystyrene (PS) matrix was achieved. The average free-volume radius in the PS/clay nanocomposites is generally same as that in PS. Along with increment of rectorite contents, the interface between rectorite and polystyrene matrix increases, and the free-volume concentration decreases obviously. And the polystyrene nanocomposites have higher thermal decomposition temperature than pure PS.

High-density Polyethylene Elucidated by Energy-tunable Positron Annihilation and Water Contact Angle Measurements Zhe Chen1,2,∗, Zhi Wang1,2, Qiuming Fu1,2, Zhibin Ma1,2, Pengfei Fang3, and Chunqing He3 1Hubei Key Laboratory of Plasma Chemistry and New Materials, Wuhan Institute of Technology, Wuhan, 430073, China 2School of Material Science and Engineering, Wuhan Institute of Technology, Wuhan, 430073, China 3School of Physics and Technology, Wuhan University, Wuhan, 430072, China

Feng Zheng1, Chunqing He1, Pengfei Fang1,∗, Zhe Chen2, Shaojie Wang1,∗, Jianguo Wang3, Xiangyang Peng4, and Zhihai Xu4 1Department of Physics and Hubei Nuclear-Solid Physics Key Laboratory, Wuhan University, Wuhan 430072, China 2Department of Material Science and Engineering, Wuhan Institute of Technology, Wuhan 430073, China 3School of Electrical Engineering, Wuhan University, Wuhan 430072, China 4Guangdong Electric Power Research Institute, Guangzhou 510080, China

The photodegradation progress of epoxy cured with polyamide and the effect of the curing agent content under UV-A irradiation have been investigated using positron annihilation spectroscopy with an energy tunable positron beam. After 88 h of irradiation, a post-cure process and the generation of carbonyl groups reduce the value of the S parameter, compared with the virgin samples. As the irradiation time increases from 208 h to 399 h, the S parameter decreases, which may be due to the growth of carbonylgroups and the generation of free radicals. After 543 h of irradiation, a dead layer with very low S value appears near the sample surface probably induced by a dramatic decrease in the Ps formation probability. The positron results also reveal that epoxy cured with an appropriate amount of polyamide has a smaller dead layer suggesting that the amount of curing agent is a key factor affecting the photodegradationof epoxy resin.

A new approach to functionalization of fullerene is developed: treating [60] fullerene p-tert-butyl calix[8]arene complex with vaporous reagents, such as chlorine, vaporous bromine or vaporous amines. A dramatic reaction acceleration was observed and derivatives with smaller substitution degree were obtained in the reaction in comparison with the reaction of [60]fullerene itself The influences of temperature and gas flow rate were investigated.

Positron annihilation lifetime spectra measurements were performed on C60-containing chitosan as a function of temperature ranging from 290 to 380 K. Both the ortho-positronium annihilation lifetime τ3 and its intensity I3 increase with increasing temperature, however, the values of I3 in C60-containing chitosan are smaller than that in pure chitosan, and the slope of τ3 with temperature is also suppressed after C60 linked. The existing three models for positron annihilation in polymer have been compared to analyze the experiment result. It is realized that only the spur reaction model, combined with the free volume model, can explain it satisfactorily. The ortho-positronium lifetime and intensity were affected by the local physicochemical environment of material, i.e., free-volume structure, electron scavenging and trapping sites, etc. The linked C60 that play a significant chemical inhibition and quenching role on positronium formation and annihilation is first observed. The microstructure change brought by C60 group has also been discussed.

Fullerenen-propylamine platinum, rhodium complex have been prepared from C60 via amination withn-propylamine, then reacting respectively with potassium chloroplatinite or rhodium chloride. The two noble metal complexes can effectively catalyze hydrosiiylation of olefins with triethoxysilane. The Pt complex exhibits high regioselectivty for styrene, nearly 100% branched product is obtained.

Fullerene-containing polysiloxanes were prepared by reacting [60]fullerene with azidoalkyl polysiloxanes directly, and characterized by FT-IR, UV-Vis, 1H-NMR and 13C-NMR. The polymers were used as stationary phases in capillary gas chromatography. Many organic compounds, especially high boiling substance or aromatic or alcohol positional isomers, were separated effectively on these columns.