The class activation maps are generated from the final convolutional layer of CNN. They can highlight discriminative object regions for the class of interest. These discovered object regions have ...been widely used for weakly-supervised tasks. However, due to the small spatial resolution of the final convolutional layer, such class activation maps often locate coarse regions of the target objects, limiting the performance of weakly-supervised tasks that need pixel-accurate object locations. Thus, we aim to generate more fine-grained object localization information from the class activation maps to locate the target objects more accurately. In this paper, by rethinking the relationships between the feature maps and their corresponding gradients, we propose a simple yet effective method, called LayerCAM. It can produce reliable class activation maps for different layers of CNN. This property enables us to collect object localization information from coarse (rough spatial localization) to fine (precise fine-grained details) levels. We further integrate them into a high-quality class activation map, where the object-related pixels can be better highlighted. To evaluate the quality of the class activation maps produced by LayerCAM, we apply them to weakly-supervised object localization and semantic segmentation. Experiments demonstrate that the class activation maps generated by our method are more effective and reliable than those by the existing attention methods. The code will be made publicly available.
Humans can naturally and effectively find salient regions in complex scenes. Motivated by this observation, attention mechanisms were introduced into computer vision with the aim of imitating this ...aspect of the human visual system. Such an attention mechanism can be regarded as a dynamic weight adjustment process based on features of the input image. Attention mechanisms have achieved great success in many visual tasks, including image classification, object detection, semantic segmentation, video understanding, image generation, 3D vision, multimodal tasks, and self-supervised learning. In this survey, we provide a comprehensive review of various attention mechanisms in computer vision and categorize them according to approach, such as channel attention, spatial attention, temporal attention, and branch attention; a related repository
https://github.com/MenghaoGuo/Awesome-Vision-Attentions
is dedicated to collecting related work. We also suggest future directions for attention mechanism research.
Carbon‐based materials are considered to be one of the most promising materials for negative electrodes of the future, because of their good chemical stability, high electrical conductivity, and ...environmental benignity. However, to date, the underlying principles of K‐ion storage in carbonaceous anodes remain elusive, which greatly hinders the development of such a category of anodes. Herein, the ultrastable K‐ion storage of carbonaceous anode through systematic analyses, including comprehensive electrochemical characterizations, kinetics calculations, and structural/compositional evolution mechanism studies, is theoretically elucidated and experimentally verified. Specifically, it is found that the uniquely envelope‐like nitrogen‐doped carbon nanosheets with high pseudocapacitive could bring ultrastable storage of potassium ions, delivering a high initial reversible capacity of 367 mAh g−1 at a current density of 50 mA g−1 and retain 70.5 and 75.6% at current densities of 500 and 1000 mA g−1 after 1000th cycle, respectively. This study could enlighten researchers on further progress in the field of carbonaceous K‐ion battery negative electrode with a long cycle life.
The ultrastable K‐ion storage of carbonaceous anode is understood by both theoretical elucidation and experimental verification. Kinetics analysis demonstrates pseudocapacitive behavior contributes significantly to stable K‐ion storage. Structural and compositional evolutions reveal high reversibility of carbonaceous structure during the charge/discharge process. Envelope‐like N‐doped carbon nanosheet exhibits an outstanding stability with 75.6% capacity remaining at 1000 mA g−1 after 1000 cycles.
Full text
Available for:
BFBNIB, FZAB, GIS, IJS, KILJ, NLZOH, NUK, OILJ, SBCE, SBMB, UL, UM, UPUK
An ultrahigh pyridinic N‐content‐doped porous carbon monolith is reported, and the content of pyridinic N reaches up to 10.1% in overall material (53.4 ± 0.9% out of 18.9 ± 0.4% N content), being ...higher than most of previously reported N‐doping carbonaceous materials, which exhibit greatly improved electrochemical performance for potassium storage, especially in term of the high reversible capacity. Remarkably, the pyridinic N‐doped porous carbon monolith (PNCM) electrode exhibits high initial charge capacity of 487 mAh g−1 at a current density of 20 mA g−1, which is one of the highest reversible capacities among all carbonaceous anodes for K‐ion batteries. Moreover, the K‐ion full cell is successfully assembled, demonstrating a high practical energy density of 153.5 Wh kg−1. These results make PNCM promising for practical application in energy storage devices and encourage more investigations on a similar potassium storage system.
An ultrahigh pyridinic N‐doped porous carbon monolith (PNCM) with a 3D interconnected structure is fabricated by using melamine–formaldehyde resin as a hard template and a nitrogen source. The electrode delivers an extremely high reversible capacity of 487 mA h g−1 at 20 mA g−1. A K‐ion full cell assembled with PNCM/3,4,9,10‐perylene tetracarboxylic acid dianhydride demonstrates promising potential for practical applications.
Full text
Available for:
BFBNIB, FZAB, GIS, IJS, KILJ, NLZOH, NUK, OILJ, SBCE, SBMB, UL, UM, UPUK
5.
Delving Deep Into Label Smoothing Zhang, Chang-Bin; Jiang, Peng-Tao; Hou, Qibin ...
IEEE transactions on image processing,
2021, Volume:
30
Journal Article
Peer reviewed
Open access
Label smoothing is an effective regularization tool for deep neural networks (DNNs), which generates soft labels by applying a weighted average between the uniform distribution and the hard label. It ...is often used to reduce the overfitting problem of training DNNs and further improve classification performance. In this paper, we aim to investigate how to generate more reliable soft labels. We present an Online Label Smoothing (OLS) strategy, which generates soft labels based on the statistics of the model prediction for the target category. The proposed OLS constructs a more reasonable probability distribution between the target categories and non-target categories to supervise DNNs. Experiments demonstrate that based on the same classification models, the proposed approach can effectively improve the classification performance on CIFAR-100, ImageNet, and fine-grained datasets. Additionally, the proposed method can significantly improve the robustness of DNN models to noisy labels compared to current label smoothing approaches. The source code is available at our project page: https://mmcheng.net/ols/
In this work, the effects of lignin on the thermal stability of bagasse cellulose nanofibrils (CNFs) were investigated. The CNFs were prepared with different lignin content bagasse pulp using ...ultrafine grinding combined with high-pressure homogenization. Scanning electron microscopy, Fourier transform infrared spectroscopy, X-ray diffraction and thermogravimetric analysis were used to study the influence mechanisms of lignin content on the thermal stability of the CNFs. The thermal stability of cellulose was tested by thermogravimetric analyzer at different heating rates, and the activation energy of bagasse cellulose nanofibrils was calculated by Flynn-Wall-Ozawa method. The results showed that the average width of CNFs prepared by the mechanical method was approximately 20 nm. The higher the lignin contents in the CNFs, the lower the crystallinity and the better the thermal stability. The thermal decomposition activation energy of CNF fluctuates with the change of conversion rate. Under the same conversion rate, the higher the lignin contents in the CNFs, the larger the activation energy value. The average activation energies of NO-LCNF, L-LCNF, ML-LCNF, MH-LCNF, and H-LCNF were 208.14, 254.49, 412.95, 530.54 and 652.10 kJ/mol, respectively, during the conversion rate of 20% to 90%. The research results provide a theoretical basis for the pyrolysis mechanism and high value utilization of CNFs and have a profound impact on promoting the application and development of CNFs in emerging nanocomposites.
Graphic abstract
Full text
Available for:
EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, KILJ, KISLJ, MFDPS, NLZOH, NUK, OBVAL, OILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
Water splitting is considered as a pollution‐free and efficient solution to produce hydrogen energy. Low‐cost and efficient electrocatalysts for the hydrogen evolution reaction (HER) and the oxygen ...evolution reaction (OER) are needed. Recently, chemical vapor deposition is used as an effective approach to gain high‐quality MoS2 nanosheets (NSs), which possess excellent performance for water splitting comparable to platinum. Herein, MoS2 NSs grown vertically on FeNi substrates are obtained with in situ growth of Fe5Ni4S8 (FNS) at the interface during the synthesis of MoS2. The synthesized MoS2/FNS/FeNi foam exhibits only 120 mV at 10 mA cm−2 for HER and exceptionally low overpotential of 204 mV to attain the same current density for OER. Density functional theory calculations further reveal that the constructed coupling interface between MoS2 and FNS facilitates the absorption of H atoms and OH groups, consequently enhancing the performances of HER and OER. Such impressive performances herald that the unique structure provides an approach for designing advanced electrocatalysts.
Strong coupling interfaces of a vertical MoS2 array and in situ grown Fe5Ni4S8 are formed by chemical vapor deposition. The interfacial coupling of the MoS2 array on FeNi foam shows outstanding activity of both the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER): 120 mV @ 10 mA cm–2 for HER and 204 mV @ 10 mA cm–2 for OER.
Full text
Available for:
BFBNIB, FZAB, GIS, IJS, KILJ, NLZOH, NUK, OILJ, SBCE, SBMB, UL, UM, UPUK
Currently, solar-thermal energy storage within phase-change materials relies on adding high thermal-conductivity fillers to improve the thermal-diffusion-based charging rate, which often leads to ...limited enhancement of charging speed and sacrificed energy storage capacity. Here we report the exploration of a magnetically enhanced photon-transport-based charging approach, which enables the dynamic tuning of the distribution of optical absorbers dispersed within phase-change materials, to simultaneously achieve fast charging rates, large phase-change enthalpy, and high solar-thermal energy conversion efficiency. Compared with conventional thermal charging, the optical charging strategy improves the charging rate by more than 270% and triples the amount of overall stored thermal energy. This superior performance results from the distinct step-by-step photon-transport charging mechanism and the increased latent heat storage through magnetic manipulation of the dynamic distribution of optical absorbers.
Surface runoff and soil loss from 2007 to 2010 related to land use and rainfall regimes in karst hill slopes in Guizhou Province, southwest China, were analyzed. Using the hierarchical clustering ...method, sixty-one rainfall events under the subtropical monsoon climate condition were classified into 5 types of rainfall regimes according to the depth, maximum 30-min intensity, and duration of rainfall. In our study, we first demonstrated that the amounts of surface runoff and soil loss on the karst hill slopes were very small compared to the non-karst areas, because the dual hydrological structure in the karst region, including ground and underground drainage systems, could influence the processes of rainfall recharge and runoff generation. Most rainfall water was transported underground through limestone fissures and fractures, while little was in the form of surface runoff. Second, the runoff and soil loss were affected by land use management and vegetation cover. Soil loss was intensified in a descending order to five types of land uses: pastureland
>
burned area
>
cropland
>
combination vegetation land
>
young forestland. Third, the runoff and soil loss exhibited remarkable variances among different rainfall regimes. Large runoff and soil loss were mainly created by heavy rainfall storms with a rainfall depth of more than 40
mm and a maximum 30-min rainfall intensity of over 30
mm
h
−
1
. In addition, rainfall storms with large antecedent precipitations could also produce large runoff and soil loss. These observations indicated that limestone fissures and fractures play important roles in surface runoff generation on karst limestone slopes due to their large storage capacity and high infiltration rate. Lastly, the soil erosion risk in the karst pure limestone slope is quite high and should be paid particular attention, especially in regards to over-grazing because the soil loss created by a single heavy rainstorm in pastureland was 5 times the annual soil loss tolerance.
► Surface runoff and soil loss on the karst hill slopes were very small. ► Runoff and soil loss were affected by land use, land cover and rainfall regimes. ► Antecedent precipitation could promote runoff and soil loss on limestone slopes. ► Due to the low soil formation rates, soil erosion risk was quite high.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UL, UM, UPCLJ, UPUK
Single-atom catalysts (SACs) have recently attracted significant attention due to their maximum atom utilization and high efficiency in a series of electrocatalytic reactions. However, the atomically ...dispersed metal atoms have intrinsically extreme mobility due to their high surface energy. Besides, the harsh reaction conditions of electrocatalysis also challenged the catalytic stability of SACs. The excellent electrocatalytic performance of SACs always degrades under long-term operating conditions. Most previous studies of SACs have focused more on the activity and selectivity of SACs in electrocatalysis, while catalytic stability has received little attention as a more critical factor limiting their large-scale industrial application. In this review, we provide an overview of the recent advances in SACs in terms of selecting the metal and support materials, synthetic strategies, and electrocatalytic applications with a focus on catalytic stability. A deep understanding of the instability behaviors of SACs under different electrocatalytic conditions contributes to the design of effective synthetic strategies to further optimize their catalytic stability, which is particularly discussed. Finally, we present the challenges and prospects for the future development of stable SACs in electrocatalysis.
The recent advances in the stability of SACs, including the selection of metals and supports, synthetic strategies, and the catalytic stability in electrocatalysis.