The problem of generalizing deep neural networks from multiple source domains to a target one is studied under two settings: When unlabeled target data is available, it is a multi-source unsupervised ...domain adaptation (UDA) problem, otherwise a domain generalization (DG) problem. We propose a unified framework termed domain adaptive ensemble learning (DAEL) to address both problems. A DAEL model is composed of a CNN feature extractor shared across domains and multiple classifier heads each trained to specialize in a particular source domain. Each such classifier is an expert to its own domain but a non-expert to others. DAEL aims to learn these experts collaboratively so that when forming an ensemble, they can leverage complementary information from each other to be more effective for an unseen target domain. To this end, each source domain is used in turn as a pseudo-target-domain with its own expert providing supervisory signal to the ensemble of non-experts learned from the other sources. To deal with unlabeled target data under the UDA setting where real expert does not exist, DAEL uses pseudo labels to supervise the ensemble learning. Extensive experiments on three multi-source UDA datasets and two DG datasets show that DAEL improves the state of the art on both problems, often by significant margins.
Face detection and alignment in unconstrained environment are challenging due to various poses, illuminations, and occlusions. Recent studies show that deep learning approaches can achieve impressive ...performance on these two tasks. In this letter, we propose a deep cascaded multitask framework that exploits the inherent correlation between detection and alignment to boost up their performance. In particular, our framework leverages a cascaded architecture with three stages of carefully designed deep convolutional networks to predict face and landmark location in a coarse-to-fine manner. In addition, we propose a new online hard sample mining strategy that further improves the performance in practice. Our method achieves superior accuracy over the state-of-the-art techniques on the challenging face detection dataset and benchmark and WIDER FACE benchmarks for face detection, and annotated facial landmarks in the wild benchmark for face alignment, while keeps real-time performance.
Recent years have witnessed the popularity of using recurrent neural network (RNN) for action recognition in videos. However, videos are of high dimensionality and contain rich human dynamics with ...various motion scales, which makes the traditional RNNs difficult to capture complex action information. In this paper, we propose a novel recurrent spatial-temporal attention network (RSTAN) to address this challenge, where we introduce a spatial-temporal attention mechanism to adaptively identify key features from the global video context for every time-step prediction of RNN. More specifically, we make three main contributions from the following aspects. First, we reinforce the classical long short-term memory (LSTM) with a novel spatial-temporal attention module. At each time step, our module can automatically learn a spatial-temporal action representation from all sampled video frames, which is compact and highly relevant to the prediction at the current step. Second, we design an attention-driven appearance-motion fusion strategy to integrate appearance and motion LSTMs into a unified framework, where LSTMs with their spatial-temporal attention modules in two streams can be jointly trained in an end-to-end fashion. Third, we develop actor-attention regularization for RSTAN, which can guide our attention mechanism to focus on the important action regions around actors. We evaluate the proposed RSTAN on the benchmark UCF101, HMDB51 and JHMDB data sets. The experimental results show that, our RSTAN outperforms other recent RNN-based approaches on UCF101 and HMDB51 as well as achieves the state-of-the-art on JHMDB.
Recent deep learning models have demonstrated strong capabilities for classifying text and non-text components in natural images. They extract a high-level feature globally computed from a whole ...image component (patch), where the cluttered background information may dominate true text features in the deep representation. This leads to less discriminative power and poorer robustness. In this paper, we present a new system for scene text detection by proposing a novel text-attentional convolutional neural network (Text-CNN) that particularly focuses on extracting text-related regions and features from the image components. We develop a new learning mechanism to train the Text-CNN with multi-level and rich supervised information, including text region mask, character label, and binary text/non-text information. The rich supervision information enables the Text-CNN with a strong capability for discriminating ambiguous texts, and also increases its robustness against complicated background components. The training process is formulated as a multi-task learning problem, where low-level supervised information greatly facilitates the main task of text/non-text classification. In addition, a powerful low-level detector called contrast-enhancement maximally stable extremal regions (MSERs) is developed, which extends the widely used MSERs by enhancing intensity contrast between text patterns and background. This allows it to detect highly challenging text patterns, resulting in a higher recall. Our approach achieved promising results on the ICDAR 2013 data set, with an F-measure of 0.82, substantially improving the state-of-the-art results.
Occlusion and pose variations, which can change facial appearance significantly, are two major obstacles for automatic Facial Expression Recognition (FER). Though automatic FER has made substantial ...progresses in the past few decades, occlusion-robust and pose-invariant issues of FER have received relatively less attention, especially in real-world scenarios. This paper addresses the real-world pose and occlusion robust FER problem in the following aspects. First, to stimulate the research of FER under real-world occlusions and variant poses, we annotate several in-the-wild FER datasets with pose and occlusion attributes for the community. Second, we propose a novel Region Attention Network (RAN), to adaptively capture the importance of facial regions for occlusion and pose variant FER. The RAN aggregates and embeds varied number of region features produced by a backbone convolutional neural network into a compact fixed-length representation. Last, inspired by the fact that facial expressions are mainly defined by facial action units, we propose a region biased loss to encourage high attention weights for the most important regions. We validate our RAN and region biased loss on both our built test datasets and four popular datasets: FERPlus, AffectNet, RAF-DB, and SFEW. Extensive experiments show that our RAN and region biased loss largely improve the performance of FER with occlusion and variant pose. Our method also achieves state-of-the-art results on FERPlus, AffectNet, RAF-DB, and SFEW. Code and the collected test data will be publicly available.
The introduction of the redox couple of triiodide/iodide (I3−/I−) into aqueous rechargeable zinc batteries is a promising energy‐storage resource owing to its safety and cost‐effectiveness. ...Nevertheless, the limited lifespan of zinc–iodine (Zn–I2) batteries is currently far from satisfactory owing to the uncontrolled shuttling of triiodide and unfavorable side‐reactions on the Zn anode. Herein, space‐resolution Raman and micro‐IR spectroscopies reveal that the Zn anode suffers from corrosion induced by both water and iodine species. Then, a metal–organic framework (MOF) is exploited as an ionic sieve membrane to simultaneously resolve these problems for Zn–I2 batteries. The multifunctional MOF membrane, first, suppresses the shuttling of I3− and restrains related parasitic side‐reaction on the Zn anode. Furthermore, by regulating the electrolyte solvation structure, the MOF channels construct a unique electrolyte structure (more aggregative ion associations than in saturated electrolyte). With the concurrent improvement on both the iodine cathode and the Zn anode, Zn–I2 batteries achieve an ultralong lifespan (>6000 cycles), high capacity retention (84.6%), and high reversibility (Coulombic efficiency: 99.65%). This work not only systematically reveals the parasitic influence of free water and iodine species to the Zn anode, but also provides an efficient strategy to develop long‐life aqueous Zn–I2 batteries.
Space‐resolution Raman and micro‐IR spectroscopies originally revealed the negative influence of water and iodine species to zinc anode. A metal–organic framework (MOF) membrane is adopted to concurrently suppress shuttling of triiodide and regulate electrolyte solvation. Benefiting from the MOF membrane, an aqueous zinc–iodide battery achieves long life (>6000 cycles), high capacity retention (84.6%), and high reversibility (Coulombic efficiency: 99.65%).
Rechargeable aqueous zinc batteries (RAZB) have been re‐evaluated because of the superiority in addressing safety and cost concerns. Nonetheless, the limited lifespan arising from dendritic ...electrodeposition of metallic Zn hinders their further development. Herein, a metal–organic framework (MOF) was constructed as front surface layer to maintain a super‐saturated electrolyte layer on the Zn anode. Raman spectroscopy indicated that the highly coordinated ion complexes migrating through the MOF channels were different from the solvation structure in bulk electrolyte. Benefiting from the unique super‐saturated front surface, symmetric Zn cells survived up to 3000 hours at 0.5 mA cm−2, near 55‐times that of bare Zn anodes. Moreover, aqueous MnO2–Zn batteries delivered a reversible capacity of 180.3 mAh g−1 and maintained a high capacity retention of 88.9 % after 600 cycles with MnO2 mass loading up to 4.2 mg cm−2.
Up front: A metal–organic framework (MOF) was constructed as a front surface layer to maintain a super‐saturated electrolyte layer on the Zn anode. Benefiting from the highly coordinated ion complexes in the MOF channels, the Zn anode underwent homogeneous zinc electrode‐position, had a lifespan of up to 3000 hours at 0.5 mA cm−2, and gave stable aqueous MnO2–Zn batteries.
Solar energy‐driven water evaporation is a promising sustainable strategy to purify seawater and contaminated water. However, developing solar evaporators with high water evaporation rates and ...excellent salt resistance still faces a great challenge. Herein, inspired by the long‐range ordered structure and water transportation capability of lotus stem, a biomimetic aerogel with vertically ordered channels and low water evaporation enthalpy for high‐efficiency solar energy‐driven salt‐resistant seawater desalination and wastewater purification is developed. The biomimetic aerogel consists of ultralong hydroxyapatite nanowires as heat‐insulating skeletons, polydopamine‐modified MXene as a photothermal material with broadband sunlight absorption and high photothermal conversion efficiency, polyacrylamide, and polyvinyl alcohol as reagents to lower the water evaporation enthalpy and as glues to enhance the mechanical performance. The honeycomb porous structure, unidirectionally aligned microchannels, and nanowire/nanosheet/polymer pore wall endow the biomimetic aerogel with excellent mechanical properties, rapid water transportation, and excellent solar water evaporation performance. The biomimetic aerogel exhibits a high water evaporation rate (2.62 kg m−2 h−1) and energy efficiency (93.6%) under one sun irradiation. The superior salt‐rejecting ability of the designed water evaporator enables stable and continuous seawater desalination, which is promising for application in water purification to mitigate the global water crisis.
A lotus stem‐inspired aerogel water evaporator with vertically aligned channels and low water evaporation enthalpy is fabricated using ultralong hydroxyapatite nanowires, polyacrylamide, polyvinyl alcohol, and polydopamine‐modified MXene, which exhibits a high water evaporation rate and high energy efficiency as well as stable and salt‐rejecting seawater desalination.
Water pollution and freshwater shortage have deteriorated the global water crisis. Developing sustainable methods to alleviate contaminated water has become an urgent affair. Herein, inspired by ...water transportation and transpiration of natural trees, the authors report an ultralong hydroxyapatite nanowires‐based biomimetic aerogel with vertically aligned channels and multiple functions for continuous flow catalysis, water disinfection, solar energy‐driven water purification, and seawater desalination. Ultralong hydroxyapatite nanowires act as carriers to immobilize catalyst nanoparticles and serve as building blocks to assemble with chitosan to form the biomimetic aerogel with structure‐function integration. Benefiting from the interconnected cellular structure, unidirectional aligned channels, nanowire‐interwoven networked pore wall, and evenly distributed catalyst nanoparticles, the biomimetic aerogel exhibits high catalytic activity (97.6% treatment efficiency) and permeability (1786 L m−2 h−1), excellent recyclability and stability in continuous flow catalytic degradation of methylene blue solely driven by gravity. The biomimetic aerogel exhibits excellent performance for bacteria removal and anti‐biofouling. The superior photothermal conversion and heat confinement properties enable the biomimetic aerogel with a high efficiency (86.7%) for solar energy‐driven seawater desalination and wastewater purification under one sun irradiation. The multifunctional biomimetic aerogel has promising applications in catalytic reactions, wastewater remediation, and environmental engineering.
A tree‐inspired multifunctional aerogel with vertically aligned channels is fabricated from ultralong hydroxyapatite nanowires, immobilized palladium nanoparticles, and chitosan through a unidirectional freeze‐drying approach. The biomimetic aerogel exhibits high performances in gravity‐driven continuous flow catalytic reactions, water disinfection, and solar energy‐driven water purification and seawater desalination.
Metal-organic frameworks (MOFs), also known as porous coordination polymers, are constructed using metal-containing nodes with organic bridges. With intrinsically ordered porous structure, MOFs are ...considered as a promising component to modify separator/electrolyte systems for rechargeable batteries. Herein, drawing upon the information collected from recent reports, we analysed and discussed the effects of MOF-based membranes on ion transmission. Both simulation calculations and experimental results indicated that the liquid electrolyte filled MOF membranes could aid in guiding the uniformity of Li ion plating, suppressing the growth of Li dendrites. More importantly, the applications of MOF-based membranes in Li-metal battery and Li-ion battery systems were also overviewed by collecting state-of-art research results. When the MOF-based membranes were used as separators, they can not only successfully restrain the adverse shuttling of electrochemical intermediate products for achieving long cycle life Li-S batteries, but can also be used to develop dual-mediator strategies for superior electrochemical performance Li-O
2
batteries and hybrid electrolyte systems for high voltage Li-ion batteries. In addition, based on the obtained progress in rechargeable lithium batteries, the potential of MOF-based membranes serving as ionic sieves in Na-metal batteries, organic redox flow batteries and liquid-anode batteries was rationally proposed. Finally, several suggestions in regard to constructing reliable batteries with MOF membranes were provided from the perspective of practical applications.
The application of electrolyte filled MOF-based membranes as ionic sieves in rechargeable batteries.