RESUMO
Visual attention advances object detection by attending neural networks to object representations. While existing methods incorporate empirical modules to empower network attention, we rethink attentive object detection from the network learning perspective in this work. We propose a NEural Attention Learning approach (NEAL) which consists of two parts. During the back-propagation of each training iteration, we first calculate the partial derivatives (a.k.a. the accumulated gradients) of the classification output with respect to the input features. We refine these partial derivatives to obtain attention response maps whose elements reflect the contributions to the final network predictions. Then, we formulate the attention response maps as extra objective functions, which are combined together with the original detection loss to train detectors in an end-to-end manner. In this way, we succeed in learning an attentive CNN model without introducing additional network structures. We apply NEAL to the two-stage object detection frameworks, which are usually composed of a CNN feature backbone, a region proposal network (RPN), and a classifier. We show that the proposed NEAL not only helps the RPN attend to objects but also enables the classifier to pay more attention to the premier positive samples. To this end, the localization (proposal generation) and classification mutually benefit from each other in our proposed method. Extensive experiments on large-scale benchmark datasets, including MS COCO 2017 and Pascal VOC 2012, demonstrate that the proposed NEAL algorithm advances the two-stage object detector over state-of-the-art approaches.
RESUMO
Compared to sufficiently etched MOFs materials, insufficiently etched MOFs materials tend to display unsatisfactory performance due to their immature structure and have been eliminated from scientific research. Herein, this work reported a novel In2S3@SnO2 heterojunction (In2S3@SnO2-HSHT) materials, which were stably synthesized in high temperature aqueous environment and equipped extraordinary photoelectrochemical (PEC) properties, fabricated by a succinct hydrothermal synthesis method using insufficiently etched MIL-68 as a self-sacrificing template. Compared with the control groups and In2S3@SnO2 heterojunctions with collapse morphology synthesized by sufficiently etched MIL-68 in high temperature aqueous environment, In2S3@SnO2-HSHT synthesized from insufficiently etched MIL-68 as a template had a massively enhanced light-harvesting capability and generated more photoinduced charge carriers due to its well-preserved hollow structure. Therefore, based on outstanding PEC performance of In2S3@SnO2-HSHT, the established PEC label-free signal-off immunosensor to detect CYFRA 21-1, revealing vivid selectivity, stability, and reproducibility. This novel strategy adopted the insufficient chemical etching method neglected by the mainstream chemical etching approaches, which solved the challenge that the stability of the sufficient etched MOFs with hollow structure cannot be maintained under the subsequent high temperature aqueous reaction conditions, and was further applied to the design of hollow heterojunction materials for photoelectrochemical fields.
Assuntos
Técnicas Biossensoriais , Técnicas Biossensoriais/métodos , Reprodutibilidade dos Testes , Técnicas Eletroquímicas/métodos , Imunoensaio/métodosRESUMO
Zero-shot learning (ZSL) tackles the novel class recognition problem by transferring semantic knowledge from seen classes to unseen ones. Semantic knowledge is typically represented by attribute descriptions shared between different classes, which act as strong priors for localizing object attributes that represent discriminative region features, enabling significant and sufficient visual-semantic interaction for advancing ZSL. Existing attention-based models have struggled to learn inferior region features in a single image by solely using unidirectional attention, which ignore the transferable and discriminative attribute localization of visual features for representing the key semantic knowledge for effective knowledge transfer in ZSL. In this paper, we propose a cross attribute-guided Transformer network, termed TransZero++, to refine visual features and learn accurate attribute localization for key semantic knowledge representations in ZSL. Specifically, TransZero++ employs an attribute â visual Transformer sub-net (AVT) and a visual â attribute Transformer sub-net (VAT) to learn attribute-based visual features and visual-based attribute features, respectively. By further introducing feature-level and prediction-level semantical collaborative losses, the two attribute-guided transformers teach each other to learn semantic-augmented visual embeddings for key semantic knowledge representations via semantical collaborative learning. Finally, the semantic-augmented visual embeddings learned by AVT and VAT are fused to conduct desirable visual-semantic interaction cooperated with class semantic vectors for ZSL classification. Extensive experiments show that TransZero++ achieves the new state-of-the-art results on three golden ZSL benchmarks and on the large-scale ImageNet dataset. The project website is available at: https://shiming-chen.github.io/TransZero-pp/TransZero-pp.html.
RESUMO
Designing efficient and stable non-precious metal catalysts remains a significant challenge for formaldehyde (HCHO) oxidation, which is an expected way to replace the employment of noble-metal catalysts. Herein, a series of atomically dispersed Co catalysts are optimized by evaporating nitrogen atoms and exploring their HCHO oxidation catalytic performance. The results show that the prepared temperature can effectively control the coordination regulation of the Co atomic site, which in turn affects the catalytic oxidation activity. Our best catalyst, the Co-N/C prepared at 1000 °C, exhibits superior activity with 92.8% of conversion at room temperature at a gas hourly space velocity (GHSV) of 72,000 mL·g-1·h-1. Extensive characterizations combined with theoretical calculations reveal that the high catalytic activity is attributed to the low-coordinated center, which can be tailored by pyrolysis temperature. This work provides an innovative strategy for catalyst design in the catalytic oxidation reaction.
RESUMO
Formaldehyde (HCHO) oxidation to improve indoor air quality has attracted extensive attention. Designing efficient catalysts for HCHO removal at room temperature still remains challenging. Herein, we report a novel strategy to boost HCHO oxidation by the synergistic effect of Pt nanoparticles and C3N4. The pyridine nitrogen of C3N4 can create Lewis base sites, which function in adsorbing and activating O2 molecules. As the preparation temperature increased, the pyridine nitrogen content increased on the C3N4 surface, leading to a more significant synergistic effect. The mechanism study by in situ DRIFTS indicated that the adsorbed O2 molecules were activated by Pt/C3N4. As a result, the Pt/C3N4-650 has the most outstanding performance for HCHO oxidation at room temperature. HCHO can be completely eliminated with a concentration of 80 ppm at room temperature at a GHSV of 50 000 ml g-1 h-1. This study will provide a new perspective to design efficient HCHO oxidation catalysts.
RESUMO
The construction of novel heterojunction is regarded as an operative scheme to promote the transport of photogenerated carriers and reduce electron-hole pair recombination to enhance the photoelectrochemical (PEC) performances. Herein, ZnCdS hollow dodecahedral nanocages (ZnCdS-HDCs) and In2S3 hollow nanorods (In2S3-HNRs), which were derived from two different of metal-organic frameworks (MOFs) by solvothermal sulfidation method and were constructed an original double-hollow heterostructure ZnCdS-HDCs@In2S3-HNRs. The intrinsic mechanism of In2S3-HNRs benefiting from unique morphology to boost the photochemical properties under visible light irradiation was illustrated. Meanwhile, the mechanism of the novel type II heterojunction with staggered matching levels was revealed, which could effectively restrict electron-hole pair reassociation separation, and accelerated charge separation and transfer. Therefore, based on the excellent PEC performance of ZnCdS- HDCs@In2S3-HNRs double-hollow heterostructure, a signal-off PEC biosensor platform without labeled was constructed for the detection of CA15-3, which manifested acceptable specificity, reproducibility and stability. Additionally, the expected PEC biosensors showed a linear response range from 1.0 × 10-5 to 10 U·mL-1 in addition to an ultralow detection limit of 3.78 × 10-6 U·mL-1. This study innovatively constructed and prepared a new double-hollow heterojunction material with superior PEC nature for the application of PEC biosensing, which exhibits a broad application prospect.
Assuntos
Técnicas Biossensoriais , Estruturas Metalorgânicas , Técnicas Eletroquímicas , Luz , Reprodutibilidade dos TestesRESUMO
Cesium lead halide perovskite nanocrystals (PNCs) are highly attractive for optoelectronic applications due to their tunable bandgap, large absorption cross section and efficient photoluminescence. However, the dynamic ligand binding and ionic lattice make PNCs extremely sensitive to polar solvents, which greatly hinders the applications of PNCs. In this work, we first synthesize ethanol-dispersed PNCs with the assistance of water using glycyrrhizic acid (GA) as the sole capping ligand. The prepared PNCs with a mean size of 14.5 nm exhibit a narrow and symmetric emission band (full width at half maximum: 18 nm) and photoluminescence (PL) quantum yield (QY) of ~38.1%. Different with the common sense, the addition of water promotes the formation of GA-passivated PNCs due to the accelerated reaction rate of precursors and the H+ dissociation of GA at presence of Lewis base water. Furthermore, the ethanol-dispersed PNCs can be further transformed into emissive ethanol gels with improved stability. Our findings provide a novel strategy to achieve stable colloidal PNCs in polar solvents.
RESUMO
INTRODUCTION: The diagnosis of chronic obstructive pulmonary disease (COPD) is based on spirometry tests that are difficult to perform in some populations. OBJECTIVES: We aimed to construct a risk assessment model using a Bayesian Network (BN) that would enable screening high-risk populations. PATIENTS AND METHODS: A provincial survey of COPD was performed with face-to-face interviews and spirometry tests among the population aged ≥40 years in Liaoning Province, northeastern China. The potential risk factors were initially identified by multivariable logistic regression, and then a BN was built. To validate its performance, cross-validation and external dataset validation were performed, and area under the curve (AUC) and accuracy of the BN were calculated. RESULTS: The estimated age-adjusted prevalence of COPD in the entire population was 21.23% (95% confidence interval [CI]: 18.35%-24.11%). The logistic regression revealed that low education level (OR=2.35, p<0.001), elderly age (OR=4.19, p<0.001), ever smoking (OR=1.49, p=0.03), lower air quality satisfaction (OR=1.55, p=0.03) were associated with COPD. For the BN, frequent cough was the strongest single risk indicator of COPD (risk=0.374). The risks increased as more factors were specified, and the top risk was 0.738, which included the combination of elderly age, smoking, wheezing during sickness, and frequent cough. The cross-validation indicated that BN performed better than logistic regression, with a mean AUC of 0.85 and the optimum accuracy of 0.87 (cutoff=0.473). CONCLUSIONS: The BN had a favorable performance in predicting COPD risks based on questionnaires. The risks associated with the combination of several risk factors should be noted.
Assuntos
Doença Pulmonar Obstrutiva Crônica , Adulto , Idoso , Teorema de Bayes , China/epidemiologia , Humanos , Doença Pulmonar Obstrutiva Crônica/diagnóstico , Doença Pulmonar Obstrutiva Crônica/epidemiologia , Medição de Risco , Inquéritos e QuestionáriosRESUMO
Siamese networks are prevalent in visual tracking because of the efficient localization. The networks take both a search patch and a target template as inputs where the target template is usually from the initial frame. Meanwhile, Siamese trackers do not update network parameters online for real-time efficiency. The fixed target template and CNN parameters make Siamese trackers not effective to capture target appearance variations. In this paper, we propose a template updating method via reinforcement learning for Siamese regression trackers. We collect a series of templates and learn to maintain them based on an actor-critic framework. Among this framework, the actor network that is trained by deep reinforcement learning effectively updates the templates based on the tracking result on each frame. Besides the target template, we update the Siamese regression tracker online to adapt to target appearance variations. The experimental results on the standard benchmarks show the effectiveness of both template and network updating. The proposed tracker SiamRTU performs favorably against state-of-the-art approaches.
RESUMO
Facilitated by deep neural networks, numerous tracking methods have made significant advances. Existing deep trackers mainly utilize independent frames to model the target appearance, while paying less attention to its temporal coherence. In this paper, we propose a recurrent memory activation network (RMAN) to exploit the untapped temporal coherence of the target appearance for visual tracking. We build the RMAN on top of the long short-term memory network (LSTM) with an additional memory activation layer. Specifically, we first use the LSTM to model the temporal changes of the target appearance. Then we selectively activate the memory blocks via the activation layer to produce a temporally coherent representation. The recurrent memory activation layer enriches the target representations from independent frames and reduces the background interference through temporal consistency. The proposed RMAN is fully differentiable and can be optimized end-to-end. To facilitate network training, we propose a temporal coherence loss together with the original binary classification loss. Extensive experimental results on standard benchmarks demonstrate that our method performs favorably against the state-of-the-art approaches.
RESUMO
In this paper, we present an end-to-end learning framework for detailed 3D face reconstruction from a single image1. Our approach uses a 3DMM-based coarse model and a displacement map in UV-space to represent a 3D face. Unlike previous work addressing the problem, our learning framework does not require supervision of surrogate ground-truth 3D models computed with traditional approaches. Instead, we utilize the input image itself as supervision during learning. In the first stage, we combine a photometric loss and a facial perceptual loss between the input face and the rendered face, to regress a 3DMM-based coarse model. In the second stage, both the input image and the regressed texture of the coarse model are unwrapped into UV-space, and then sent through an image-toimage translation network to predict a displacement map in UVspace. The displacement map and the coarse model are used to render a final detailed face, which again can be compared with the original input image to serve as a photometric loss for the second stage. The advantage of learning displacement map in UV-space is that face alignment can be explicitly done during the unwrapping, thus facial details are easier to learn from large amount of data. Extensive experiments demonstrate the superiority of the proposed method over previous work.
RESUMO
Correlation filters (CF) have received considerable attention in visual tracking because of their computational efficiency. Leveraging deep features via off-the-shelf CNN models (e.g., VGG), CF trackers achieve state-of-the-art performance while consuming a large number of computing resources. This limits deep CF trackers to be deployed to many mobile platforms on which only a single-core CPU is available. In this paper, we propose to jointly compress and transfer off-the-shelf CNN models within a knowledge distillation framework. We formulate a CNN model pretrained from the image classification task as a teacher network, and distill this teacher network into a lightweight student network as the feature extractor to speed up CF trackers. In the distillation process, we propose a fidelity loss to enable the student network to maintain the representation capability of the teacher network. Meanwhile, we design a tracking loss to adapt the objective of the student network from object recognition to visual tracking. The distillation process is performed offline on multiple layers and adaptively updates the student network using a background-aware online learning scheme. The online adaptation stage exploits the background contents to improve the feature discrimination of the student network. Extensive experiments on six standard datasets demonstrate that the lightweight student network accelerates the speed of state-of-the-art deep CF trackers to real-time on a single-core CPU while maintaining almost the same tracking accuracy.
RESUMO
Herein, novel photoactive materials, MOF-derived porous hollow carbon nanobubbles@ZnCdS multi-shelled dodecahedral cages (C@ZnCdS MSDCs)ï¼ were synthesized via continuous chemical etching, sulfurization, cation-exchange and calcination strategies. Due to the synergistic effect between the porous shells and the carbon-layer coatingï¼ C@ZnCdS MSDCs displayed superior photoelectrochemical (PEC) performance. The synthesized C@ZnCdS MSDCs were assembled onto TiO2 modified ITO electrodes to form a type-II heterostructures. Then, Au nanoparticles (NPs) were deposited on the surface of ITO/TiO2/C@ZnCdS MSDCs. Benefiting from the unique structure and performance merits of photoactive materials, a label-free PEC sensing platform based on ITO/TiO2/C@ZnCdS MSDCs/Au was successfully constructed for CEA detection. Under optimal conditions, the PEC biosensor exhibited a wide linear range (0.00005-500â¯ngâ¯mL-1) and low detection limit (2.28â¯fgâ¯mL-1). The proposed PEC biosensor also showed good stability, specificity, reproducibility and acceptability in human serum. The prepared C@ZnCdS MSDCs would be a promising photoactive material for PEC biosensors. Most importantly, this work opens up new horizons for the application of MOFs-derived hollow carbon materials in sensing.
Assuntos
Técnicas Biossensoriais , Antígeno Carcinoembrionário/isolamento & purificação , Técnicas Eletroquímicas , Nanopartículas Metálicas/química , Cádmio/química , Antígeno Carcinoembrionário/sangue , Ouro/química , Humanos , Luz , Limite de Detecção , Porosidade , Titânio/química , Zinco/químicaRESUMO
The tracking-by-detection framework receives growing attention through the integration with the convolutional neural networks (CNNs). Existing tracking-by-detection-based methods, however, fail to track objects with severe appearance variations. This is because the traditional convolutional operation is performed on fixed grids, and thus may not be able to find the correct response while the object is changing pose or under varying environmental conditions. In this paper, we propose a deformable convolution layer to enrich the target appearance representations in the tracking-by-detection framework. We aim to capture the target appearance variations via deformable convolution, which adaptively enhances its original features. In addition, we also propose a gated fusion scheme to control how the variations captured by the deformable convolution affect the original appearance. The enriched feature representation through deformable convolution facilitates the discrimination of the CNN classifier on the target object and background. The extensive experiments on the standard benchmarks show that the proposed tracker performs favorably against the state-of-the-art methods.
RESUMO
Background: This network meta-analysis aimed at comparing anti-programmed death 1 (anti-PD-1) with anti-programmed death ligand 1(anti-PD-L1) immunotherapy in patients with metastatic, previously treated non-small cell lung cancer (NSCLC) who failed first-line treatment. Methods: We searched electronic databases to identify all eligible clinical trials. End-points included overall survival (OS), progression-free survival (PFS) and objective response. Hazard ratios (HRs) or odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were extracted. Network meta-analysis was performed using the frequentist approach for multiple treatment comparisons. Results: In total, 3024 patients were randomly assigned: 1117 received anti-PD-1 therapy (nivolumab + pembrolizumab), 569 received anti-PD-L1 (atezolizumab) and 1338 received docetaxel. Anti-PD-1 (HR, 0.56; 95% CI, 0.48-0.66) and anti-PD-L1 (HR, 0.64; 95% CI, 0.51-0.79) achieved better OS than docetaxel, and anti-PD-1 was superior to docetaxel in terms of PFS (HR, 0.75; 95% CI, 0.62-0.89). Moreover, anti-PD-1 achieved the highest effect on OS and PFS, with a P-score of 91.2% and 95.5%, respectively. With regard to tumor response, anti-PD-1 group had a higher rate of responders than that in anti-PD-L1 (HR, 0.35; 95% CI, 0.19-0.65) and docetaxel (HR, 0.36; 95% CI, 0.25-0.52) groups. Undoubtedly, anti-PD-1 and anti-PD-L1 obtained less toxicity profile than docetaxel, and no significant difference was observed between anti-PD-1 and anti-PD-L1 groups. Conclusions: Anti-PD-1 may be a better choice for patients with metastatic and previously treated NSCLC who failed first-line treatment in terms of the treatment ranking.
RESUMO
Reduced graphite oxide (rGO) was incorporated into a metal organic framework (MOF) MIL-101(Cr) for the modification of carbon paste electrode. Taking advantages of the large surface area of MOF and the electrical conductivity of rGO, the resulted electrodes exhibited high sensitivity and reliability in the simultaneous electrochemical identification and quantification of catechol (CC) and hydroquinone (HQ). Specifically, in the mixture solution of catechol and hydroquinone (constant concentration of an analyte), the linear response ranges for catechol and hydroquinone were 10-1400⯵M and 4-1000⯵M, and detection limits were 4⯵M and 0.66 µM (S/Nâ¯=â¯3) for individual catechol and hydroquinone, respectively. Therefore, the relatively easy fabrication of modified CPE and its fascinating reliability towards HQ and CC detection may simulate more research interest in the applications of MIL-101(Cr)-rGO composites for electrochemical sensors.
RESUMO
Background and Objective: Both induction chemotherapy (IC) followed by concurrent chemoradiotherapy (CCRT; IC+CCRT) and CCRT plus adjuvant chemotherapy (AC; CCRT+AC) are standard treatments for advanced nasopharyngeal carcinoma (NPC). However, no prospective randomized trials comparing these two approaches have been published yet. We conducted this network meta-analysis to address this clinical question. Method: We recruited randomized clinical trials involving patients with advanced NPC randomly allocated to IC+CCRT, CCRT+AC, CCRT, or radiotherapy (RT) alone. Pairwise meta-analysis was first conducted, then network meta-analysis was performed using the frequentist approach. Effect size was expressed as hazard ratio (HR) and 95% confidence interval (CI). Results: Overall, 12 trials involving 3,248 patients were recruited for this study, with 555 receiving IC+CCRT, 840 receiving CCRT+AC, 1,039 receiving CCRT, and 814 receiving radiotherapy (RT) alone. IC+CCRT achieved significantly better overall survival ([HR], 0.69; 95% [CI], 0.51-0.92), distant metastasis-free survival (HR, 0.58; 95% CI, 0.44-0.78), and locoregional recurrence-free survival (HR, 0.67; 95% CI, 0.47-0.98) than CCRT. However, survival outcomes did not significantly differ between IC+CCRT and CCRT+AC, or between CCRT+AC and CCRT arms for all the endpoints. As expected, RT alone is the poorest treatment. In terms of P-score, IC+CCRT ranked best for overall survival (96.1%), distant metastasis-free survival (99.0%) and locoregional recurrence-free survival (87.1%). Conclusions: IC+CCRT may be a better and more promising treatment strategy for advanced NPC; however, head-to-head randomized trials comparing IC-CCRT with CCRT-AC are warranted.
RESUMO
The adsorption capacities of a heterometallic metal-organic framework (CPM-200-In/Mg) to VOCs (HCHO, C2H4, CH4, C2H2, C3H8, C2H6, C2H3Cl, C2H2Cl2, CH2Cl2 and CHCl3) and some inorganic gas molecules (HCN, SO2, NO, CO2, CO, H2S and NH3), as well as its selectivity in ternary mixture systems of natural gas and post-combustion flue gas are theoretically explored at the grand canonical Monte Carlo (GCMC) and density functional theory (DFT) levels. It is shown that CPM-200-In/Mg is suitable for the adsorption of VOCs, particularly for HCHO (up to 0.39 g g-1 at 298 K and 1 bar), and the adsorption capacities of some inorganic gas molecules such as SO2, H2S and CO2 match well with the sequence of their polarizability (SO2 > H2S > CO2). The large adsorption capacities of HCN and HCHO in the framework result from the strong interaction between adsorbates and metal centers, based on analyzing the radial distribution functions (RDF). Comparing C2H4 and CH4 molecules interacting with CPM-200-In/Mg by VDW interaction, we speculate that the high adsorption capacities of their chlorine derivatives in the framework could be due to the existence of halogen bonding or strong electrostatic and VDW interactions. It is found that the basic groups, including -NH2, -N and -OH, can effectively improve both the adsorption capacities and selectivity of CPM-200-In/Mg for harmful gases. Note that the adsorption capacity of CPM-200-In/Mg-NH2 (site 2) (245 cm3 g-1) for CO2 exceeded that of MOF-74-Mg (228 cm3 g-1) at 273 K and 1 bar and that for HCHO can reach 0.41 g g-1, which is almost twice that of 438-MOF and nearly 45 times of that in active carbon. Moreover, for natural gas mixtures, the decarburization and desulfurization abilities of CPM-200-In/Mg-NH2 (site 2) have exceeded those of the MOF-74 series, while for post-combustion flue gas mixtures, the desulfurization ability of CPM-200-In/Mg-NH2 (site 2) is still comparable to those of the MOF-74 series at 303 K and 4 MPa. We hope that the current theoretical study could guide experimental research in the future.
RESUMO
We present a new efficient edge-preserving filter-"tree filter"-to achieve strong image smoothing. The proposed filter can smooth out high-contrast details while preserving major edges, which is not achievable for bilateral-filter-like techniques. Tree filter is a weighted-average filter, whose kernel is derived by viewing pixel affinity in a probabilistic framework simultaneously considering pixel spatial distance, color/intensity difference, as well as connectedness. Pixel connectedness is acquired by treating pixels as nodes in a minimum spanning tree (MST) extracted from the image. The fact that an MST makes all image pixels connected through the tree endues the filter with the power to smooth out high-contrast, fine-scale details while preserving major image structures, since pixels in small isolated region will be closely connected to surrounding majority pixels through the tree, while pixels inside large homogeneous region will be automatically dragged away from pixels outside the region. The tree filter can be separated into two other filters, both of which turn out to have fast algorithms. We also propose an efficient linear time MST extraction algorithm to further improve the whole filtering speed. The algorithms give tree filter a great advantage in low computational complexity (linear to number of image pixels) and fast speed: it can process a 1-megapixel 8-bit image at ~ 0.25 s on an Intel 3.4 GHz Core i7 CPU (including the construction of MST). The proposed tree filter is demonstrated on a variety of applications.
RESUMO
The sample was obtained from original sample using "sawing method" and was dissolved with a mixed solvent of nitric acid and tartaric acid. After the sample solution was pretreated with a series of methods, Sb and Se were determined by inductively coupled plasma atomic emission spectrometry (ICP-AES). The results show that the recovery of standard addition of Sb was 96.2% and that of Se was 92.1%, and that the relative standard deviations (n=11) were smaller than 3.56%. The method is efficient, accurate and easy to operate, and has been applied to the determination of Sb and Se in Pb-Sb alloy products with satisfactory results.