RESUMO
Despite significant progress achieved in artificial self-sorting in solution, operating self-sorting in the body remains a considerable challenge. Here, we report an in vivo self-sorting peptide system via an in situ assembly evolution for combined cancer therapy. The peptide E3C16-SS-EIY consists of two disulfide-connected segments, E3C16SH and SHEIY, capable of independent assembly into twisted or flat nanoribbons. While E3C16-SS-EIY assembles into nanorods, exposure to glutathione (GSH) leads to the conversion of the peptide into E3C16SH and SHEIY, thus promoting in situ evolution from the nanorods into self-sorted nanoribbons. Furthermore, incorporation of two ligand moieties targeting antiapoptotic protein XIAP and organellar endoplasmic reticulum (ER) into the self-sorted nanoribbons allows for simultaneous inhibition of XIAP and accumulation surrounding ER. This leads to the cytotoxicity toward the cancer cells with elevated GSH levels, through activating caspase-dependent apoptosis and inducing ER dysfunction. In vivo self-sorting of E3C16-SS-EIY decorated with ligand moieties is thoroughly validated by tissue studies. Tumor-bearing mouse experiments confirm the therapeutic efficacy of the self-sorted assemblies for inhibiting tumor growth, with excellent biosafety. Our findings demonstrate an efficient approach to develop in vivo self-sorting systems and thereby facilitating in situ formulation of biomedical agents.
Assuntos
Peptídeos , Humanos , Animais , Peptídeos/química , Peptídeos/farmacologia , Camundongos , Antineoplásicos/farmacologia , Antineoplásicos/química , Apoptose/efeitos dos fármacos , Proteínas Inibidoras de Apoptose Ligadas ao Cromossomo X/antagonistas & inibidores , Proteínas Inibidoras de Apoptose Ligadas ao Cromossomo X/metabolismo , Glutationa/química , Glutationa/metabolismo , Linhagem Celular Tumoral , Nanotubos/químicaRESUMO
Nanozymes, with their versatile composition and structural adaptability, present distinct advantages over natural enzymes including heightened stability, customizable catalytic activity, cost-effectiveness, and simplified synthesis process, making them as promising alternatives in various applications. Recent advancements in nanozyme research have shifted focus from serendipitous discovery toward a more systematic approach, leveraging machine learning, theoretical calculations, and mechanistic explorations to engineer nanomaterial structures with tailored catalytic functions. Despite its pivotal role, electron transfer, a fundamental process in catalysis, has often been overlooked in previous reviews. This review comprehensively summarizes recent strategies for modulating electron transfer processes to fine-tune the catalytic activity and specificity of nanozymes, including electron-hole separation and carrier transfer. Furthermore, the bioapplications of these engineered nanozymes, including antimicrobial treatments, cancer therapy, and biosensing are also introduced. Ultimately, this review aims to offer invaluable insights for the design and synthesis of nanozymes with enhanced performance, thereby advancing the field of nanozyme research.
RESUMO
Considering the impact of differences in watershed characteristics on river water quality, with the Chaohu Lake Basin as the research object, based on the data of water quality, meteorology, topography, soil, and remote sensing images of the river monitoring points from October 2019 to September 2020, the watershed unit at each monitoring point was divided through digital terrain analysis, and the comprehensive landscape characteristics based on the watershed unit were explored through the comprehensive use of correlation analysis, redundancy analysis, and multiple regression analysis to investigate the influence of comprehensive landscape characteristics based on watershed units (including land use, climate, topography, soil, etc.) on the water quality of rivers around Chaohu Lake. The results showed that:â the water quality of rivers around Chaohu Lake had large spatial differences, with the main pollutants being total nitrogen and ammonia nitrogen. Most of the rivers had total nitrogen concentrations exceeding the Class V water quality standards, and the areas with serious nitrogen and phosphorus pollution were concentrated in the urban area of Hefei and the surrounding rivers, as well as in the middle and lower reaches of the Fengle and Hangbu Rivers. â¡ The comprehensive landscape characteristics of the watershed unit had a significant impact on the river water quality. Among them, the proportion of built-up land, the density of patches, the dispersion and juxtaposition index, and the Shannon diversity index were positively correlated with the water quality indicators, whereas the proportion of forest and grassland and the spreading index were negatively correlated with the water quality indicators. ⢠In different seasons, the effect of the integrated landscape characteristics of the watershed unit on river water quality was stronger in the wet season than in the dry season, which was mainly caused by the difference in precipitation in the dry and wet seasons.
RESUMO
Quantum dots (QDs) exhibit superior brightness and photochemical stability, making them the preferred option for highly sensitive single-molecule detection compared with fluorescent dyes or proteins. Nevertheless, their high surface energy leads to nonspecific adsorption and poor colloidal stability. In the past decades, we have found that QD-based fluorescent nanoparticles (FNs) can not only address these limitations but also enhance detection sensitivity. However, the photoluminescence quantum yield (PLQY) of FNs is significantly lower compared with that of original QDs. It is urgent to develop a strategy to solve the issue, aiming to further enhance detection sensitivity. In this study, we found that the decrease of PLQY of FNs prepared by free radical polymerization was attributed to two factors: (1) generation of defects that can cause nonradiative transitions resulting from QD-ligands desorption and QD-shell oxidation induced by free radicals; (2) self-absorption resulting from aggregation caused by incompatibility of QDs with polymers. Based on these, we proposed a multihierarchical regulation strategy that includes: (1) regulating QD-ligands; (2) precisely controlling free radical concentration; and (3) constructing cross-linked structures of polymer to improve compatibility and to reduce the formation of surface defects. It is crucial to emphasize that the simultaneous coordination of multiple factors is essential. Consequently, a world-record PLQY of 97.6% for FNs was achieved, breaking through the current bottleneck at 65%. The flexible application of this regulatory concept paves the way for the large-scale production of high-brightness QD-polymer complexes, enhancing their potential applications in sensitive biomedical detection.
RESUMO
Here, we develop a novel methodology for synthesizing chiral CdSe@ZnS quantum dots (QDs) with enhanced circularly polarized luminescence (CPL) by incorporating l-/d-histidine (l-/d-His) ligands during ZnS shell growth at the water/oil interface. The resulting chiral QDs exhibit exceptional absolute photoluminescence quantum yield of up to 67.2%, surpassing the reported limits of 40.0% for chiral inorganic QDs, along with absorption dissymmetry factor (|gabs|) and luminescence dissymmetry factor (|glum|) values of 10-2, exceeding the range of 10-5-10-3 and 10-4-10-2, respectively. Detailed investigations of the synthetic pathway reveal that the interface, as a binary synthetic environment, facilitates the coordinated ligand exchange and shell growth mediated by chiral His-Zn2+ coordination complexes, leading to a maximum fluorescent brightness and chiroptical activities. The growth process, regulated by the His-Zn2+ coordination complex, not only reduces trap states on the CdSe surface, thereby enhancing the fluorescence intensity, but also significantly promotes the orbital hybridization between QDs and chiral ligands, effectively overcoming the shielding effect of the wide bandgap shell and imparting pronounced chirality. The proposed growth pathway elucidates the origin of chirality and provides insights into the regulation of the CPL intensity in chiral QDs. Furthermore, the application of CPL QDs in multilevel anticounterfeiting systems overcomes the limitations of replication in achiral fluorescence materials and enhances the system's resistance to counterfeiting, thus opening new opportunities for chiral QDs in optical anticounterfeiting and intelligent information encryption.
RESUMO
With the development of near-infrared II (NIR-II) fluorescence imaging, Ag2Se quantum dots (QDs) have become promising label candidates due to their negligible toxicity and narrow band gap. Despite their potential for gastrointestinal (GI) imaging, the application of Ag2Se QDs still presents significant challenges due to issues such as fluorescence extinction or poor stability in the complex digestive microenvironment. Herein, we have proposed a novel approach to the continuous production of Se precursors using glutathione (GSH) as the reductant under acidic conditions, realizing the continuous growth of water-dispersible Ag2Se QDs. The Ag2Se QDs emitting at 600-1100 nm have been successfully synthesized. Meanwhile, the silver-rich surface of the synthesized NIR-II Ag2Se QDs has been passivated well with the dense GSH, resulting in exceptional colloidal stability and photostability and endowing them with acid resistance. As a result, the obtained NIR-II Ag2Se QDs have exhibited remarkable stability in gastric acid, thus enabling their utilization for long-term real-time monitoring of GI peristalsis via NIR-II fluorescence imaging. Moreover, in contrast to conventional barium meal-based X-ray imaging, NIR-II fluorescence imaging with as-prepared NIR-II Ag2Se QDs can offer clearer visualization of fine intestinal structures, with a width as small as 1.07 mm. The developed strategy has offered a new opportunity for the synthesis of acid-resistant nanocrystals, and the acid-resistant, low-toxicity, and biocompatible NIR-II Ag2Se QDs synthesized in this work show a great promise for GI imaging and diagnosis of GI diseases in vivo.
Assuntos
Nanopartículas , Pontos Quânticos , Pontos Quânticos/toxicidade , Pontos Quânticos/química , Nanopartículas/química , Fluorescência , Prata/químicaRESUMO
Clusters are considered to become increasingly significant for elaborating the nanocrystal's formation mechanism. However, capturing the clusters with high chemical potential is challenging because of the lack of effective strategies. In this work, the key role of ligand-solvent interaction has been revealed for the stabilization of clusters in silver telluride synthesis. The Flory interaction coefficient that comprehensively regards the temperature and dispersion, polarity, and hydrogen bonding of the solvent has been used to evaluate the ligand-solvent interaction and thus assist in the design of synthetic systems. Small silver telluride clusters have been successfully captured, and the composition of the smallest cluster is determined as Ag7Te8(SCy)2 (SCy represents the ligand). This work provides new insights into the design of cluster/nanocrystal synthesis systems and paves the way to revealing the mechanism of precursor-cluster-nanocrystal conversion.
RESUMO
OBJECTIVES: This study aimed to develop a liver metastasis-related gene prognostic index (LMPI) for pancreatic ductal adenocarcinoma prognosis and therapy. METHODS: The Cancer Genome Atlas data set was used to identify liver metastasis-related hub genes via weighted gene coexpression network analysis. The core genes were identified to construct an LMPI by using the Cox regression method. An immune cell abundance identifier was applied to determine the immune cell abundance. RESULTS: A total of 78 hub liver metastasis-related genes in the black module were significantly enriched in complement and coagulation cascades, fat digestion and absorption, and the PPAR signaling pathway. Then, an LMPI was constructed on the basis of the 5 prognostic genes (MOGAT3, ASGR1, TRPM8, SGSM1, and LOC101927851). Patients with higher LMPI scores had poor overall survival, more co-occurring or mutually exclusive pairs of driver gene mutations, and less benefit from immunotherapy than patients with lower LMPI scores. In addition, a high correlation was also found between LMPI scores and immune infiltration, such as CD4 naive, CD8 T, cytotoxic T, T helper 2, follicular helper T, and natural killer cells. CONCLUSIONS: The core genes of the LMPI developed may be independent factors for predicting prognosis, immune characteristics, and immunotherapy efficacy in pancreatic ductal adenocarcinoma.
Assuntos
Carcinoma Ductal Pancreático , Neoplasias Hepáticas , Neoplasias Pancreáticas , Humanos , Neoplasias Pancreáticas/genética , Neoplasias Pancreáticas/terapia , Carcinoma Ductal Pancreático/genética , Carcinoma Ductal Pancreático/terapia , Neoplasias Hepáticas/genética , Neoplasias Hepáticas/terapia , Prognóstico , Receptor de Asialoglicoproteína , Neoplasias PancreáticasRESUMO
Epidermal growth factor receptor (EGFR)-mutant lung adenocarcinoma (LUAD) patients often respond to EGFR tyrosine kinase inhibitors (TKIs) initially but eventually develop resistance to TKIs. The switch of EGFR downstream signaling from TKI-sensitive to TKI-insensitive is a critical mechanism-driving resistance to TKIs. Identification of potential therapies to target EGFR effectively is a potential strategy to treat TKI-resistant LUADs. In this study, we developed a small molecule diarylheptanoid 35d, a curcumin derivative, that effectively suppressed EGFR protein expression, killed multiple TKI-resistant LUAD cells in vitro, and suppressed tumor growth of EGFR-mutant LUAD xenografts with variant TKI-resistant mechanisms including EGFR C797S mutations in vivo. Mechanically, 35d triggers heat shock protein 70-mediated lysosomal pathway through transcriptional activation of several components in the pathway, such as HSPA1B, to induce EGFR protein degradation. Interestingly, higher HSPA1B expression in LUAD tumors associated with longer survival of EGFR-mutant, TKI-treated patients, suggesting the role of HSPA1B on retarding TKI resistance and providing a rationale for combining 35d with EGFR TKIs. Our data showed that combination of 35d significantly inhibits tumor reprogression on osimertinib and prolongs mice survival. Overall, our results suggest 35d as a promising lead compound to suppress EGFR expression and provide important insights into the development of combination therapies for TKI-resistant LUADs, which could have translational potential for the treatment of this deadly disease.
Assuntos
Adenocarcinoma de Pulmão , Diarileptanoides , Resistencia a Medicamentos Antineoplásicos , Neoplasias Pulmonares , Animais , Humanos , Camundongos , Adenocarcinoma de Pulmão/tratamento farmacológico , Adenocarcinoma de Pulmão/genética , Linhagem Celular Tumoral , Diarileptanoides/farmacologia , Receptores ErbB/metabolismo , Neoplasias Pulmonares/tratamento farmacológico , Neoplasias Pulmonares/genética , Neoplasias Pulmonares/metabolismo , Lisossomos/metabolismo , Mutação , Inibidores de Proteínas Quinases/farmacologiaRESUMO
The redox homeostasis in tumors enhances their antioxidant defense ability, limiting reactive oxygen species mediated tumor therapy efficacy. The development of strategies for specific and continuous disruption of the redox homeostasis in tumor cells facilitates the improvement of the cancer therapeutic effect by promoting the apoptosis of tumor cells. Herein, a responsively biodegradable targeting multifunctional integrated nanosphere (HDMn-QDs/PEG-FA) is designed to enhance the anti-tumor efficacy by triggering intratumoral cascade reactions to effectively disrupt intracellular redox homeostasis. Once HDMn-QDs/PEG-FA enters tumor cells, manganese dioxide (MnO2 ) shell on the surface of nanosphere consumes glutathione (GSH) to produce Mn2+ , enabling enhanced chemodynamic therapy (CDT) via a Fenton-like reaction and T1 -weighted magnetic resonance imaging. Meanwhile, the degradation of MnO2 can also cause the fluorescence recovery of quantum dots conjugated on the surface of the shell, realizing "turn-on" fluorescence imaging. In addition, the doxorubicin is released because of the cleavage of the embedded SS bond in the hybrid core framework by GSH. A superior synergistic therapeutic efficiency combined CDT and chemotherapy is shown by HDMn-QDs/PEG-FA in vivo. The tumor-inhibition rate reaches to 94.8% and does not cause normal tissue damage due to the good targeting and tumor microenvironment-specific response.
Assuntos
Nanopartículas , Nanosferas , Neoplasias , Humanos , Linhagem Celular Tumoral , Glutationa/química , Peróxido de Hidrogênio/metabolismo , Compostos de Manganês/química , Nanopartículas/química , Neoplasias/diagnóstico por imagem , Neoplasias/tratamento farmacológico , Oxirredução , Óxidos/química , Microambiente TumoralRESUMO
Live cells, as reservoirs of biochemical reactions, can serve as amazing integrated chemical plants where precursor formation, nucleation and growth of nanocrystals, and functional assembly, can be carried out accurately following an artificial program. It is crucial but challenging to deliberately direct intracellular pathways to synthesize desired nanocrystals that cannot be produced naturally in cells, because the relevant reactions exist in different spatiotemporal dimensions and will never encounter each other spontaneously. This article summarizes the progress in the introduction of inorganic functional nanocrystals into live cells via the 'artificially regulated space-time-coupled live-cell synthesis' strategy. We also describe ingenious bio-applications of nanocrystal-cell systems, and quasi-biosynthesis strategies expanded from live-cell synthesis. Artificially regulated live-cell synthesis-which involves the interdisciplinary application of biology, chemistry, nanoscience and medicine-will enable researchers to better exploit the unanticipated potentialities of live cells and open up new directions in synthetic biology.
RESUMO
Bacterial infectious diseases are common clinical diseases that seriously threaten human health, especially in countries and regions with poor environmental hygiene. Due to the lack of characteristic clinical symptoms and signs, it is a challenge to distinguish a bacterial infection from other infections, leading to misdiagnosis and antibiotic overuse. Therefore, there is an urgent need to develop a specific method for detection of bacterial infections. Herein, utilizing ultrabright fluorescent nanospheres (FNs) as reporters, immunochromatographic dyad test strips are developed for the early detection of bacterial infections and distinction of different stages of bacterial infectious diseases in clinical samples. C-reactive protein (CRP) and heparin-binding protein (HBP) are quantified and assayed because their levels in plasma are varied dynamically and asynchronously during the progression of the disease. The detection limits of CRP and HBP can reach as low as 0.51 and 0.65 ng/mL, respectively, due to the superior fluorescence intensity of each FN, which is 570 times stronger than that of a single quantum dot. The assay procedure can be achieved in 22 min, fully meeting the needs of rapid and ultrasensitive detection in the field. This constructed strip has been successfully used to profile the stage and severity of bacterial infections by monitoring the levels of CRP and HBP in human plasma samples, showing great potential as a point-of-care biosensor for clinical diagnosis. In addition to bacterial infections, the developed ultrabright FN-based point-of-care testing can be readily expanded for rapid, quantitative, and ultrasensitive detection of other trace substances in complex systems.
Assuntos
Infecções Bacterianas , Técnicas Biossensoriais , Doenças Transmissíveis , Nanosferas , Pontos Quânticos , Infecções Bacterianas/diagnóstico , Técnicas Biossensoriais/métodos , Proteína C-Reativa/análise , Humanos , Nanosferas/química , Pontos Quânticos/químicaRESUMO
Self-sorting is a common phenomenon in eukaryotic cells and represents one of the versatile strategies for the formation of advanced functional materials; however, developing artificial self-sorting assemblies within living cells remains challenging. Here, we report on the GSH-responsive in situ self-sorting peptide assemblies within cancer cells for simultaneous organelle targeting to promote combinatorial organelle dysfunction and thereby cell death. The self-sorting system was created via the design of two peptides E3C16E6 and EVMSeO derived from lipid-inspired peptide interdigitating amphiphiles and peptide bola-amphiphiles, respectively. The distinct organization patterns of the two peptides facilitate their GSH-induced self-sorting into isolated nanofibrils as a result of cleavage of disulfide-connected hydrophilic domains or reduction of selenoxide groups. The GSH-responsive in situ self-sorting in the peptide assemblies within HeLa cells was directly characterized by super-resolution structured illumination microscopy. Incorporation of the thiol and ER-targeting groups into the self-sorted assemblies endows their simultaneous targeting of endoplasmic reticulum and Golgi apparatus, thus leading to combinatorial organelle dysfunction and cell death. Our results demonstrate the establishment of the in situ self-sorting peptide assemblies within living cells, thus providing a unique platform for drug targeting delivery and an alternative strategy for modulating biological processes in the future.
Assuntos
Complexo de Golgi , Peptídeos , Retículo Endoplasmático/metabolismo , Complexo de Golgi/metabolismo , Células HeLa , Humanos , Peptídeos/química , Transporte ProteicoRESUMO
The formation of atherosclerotic plaques is the root cause of various cardiovascular diseases (CVDs). Effective CVD interventions thus call for precise identification of the plaques to aid clinical assessment, diagnosis, and treatment of such diseases. In this study, we introduce a dual-target sequentially activated fluorescence reporting system, termed in-sequence high-specificity dual-reporter unlocking (iSHERLOCK), to precisely identify the atherosclerotic plaques in vivo and ex vivo. ISHERLOCK was achieved by creating a three-in-one fluorescent probe that permits highly specific and sensitive detection of lipid droplets and hypochlorous acid via "off-on" and ratiometric readouts, respectively. Based on this format, the upregulated lipid accumulation and oxidative stress-the two hallmarks of atherosclerosis (AS)-were specifically measured in the atherosclerotic plaques, breaking through the barrier of precise tissue biopsy of AS and thus aiding effective CVD stewardship.
Assuntos
Aterosclerose , Placa Aterosclerótica , Aterosclerose/diagnóstico por imagem , Fluorescência , Corantes Fluorescentes , Humanos , Placa Aterosclerótica/diagnóstico por imagem , Placa Aterosclerótica/patologiaRESUMO
Recent advances in 3-D sensors and 3-D modeling have led to the availability of massive amounts of 3-D data. It is too onerous and time consuming to manually label a plentiful of 3-D objects in real applications. In this article, we address this issue by transferring the knowledge from the existing labeled data (e.g., the annotated 2-D images or 3-D objects) to the unlabeled 3-D objects. Specifically, we propose a domain-adversarial guided siamese network (DAGSN) for unsupervised cross-domain 3-D object retrieval (CD3DOR). It is mainly composed of three key modules: 1) siamese network-based visual feature learning; 2) mutual information (MI)-based feature enhancement; and 3) conditional domain classifier-based feature adaptation. First, we design a siamese network to encode both 3-D objects and 2-D images from two domains because of its balanced accuracy and efficiency. Besides, it can guarantee the same transformation applied to both domains, which is crucial for the positive domain shift. The core issue for the retrieval task is to improve the capability of feature abstraction, but the previous CD3DOR approaches merely focus on how to eliminate the domain shift. We solve this problem by maximizing the MI between the input 3-D object or 2-D image data and the high-level feature in the second module. To eliminate the domain shift, we design a conditional domain classifier, which can exploit multiplicative interactions between the features and predictive labels, to enforce the joint alignment in both feature level and category level. Consequently, the network can generate domain-invariant yet discriminative features for both domains, which is essential for CD3DOR. Extensive experiments on two protocols, including the cross-dataset 3-D object retrieval protocol (3-D to 3-D) on PSB/NTU, and the cross-modal 3-D object retrieval protocol (2-D to 3-D) on MI3DOR-2, demonstrate that the proposed DAGSN can significantly outperform state-of-the-art CD3DOR methods.
RESUMO
In this article, we propose a novel deep correlated joint network (DCJN) approach for 2-D image-based 3-D model retrieval. First, the proposed method can jointly learn two distinct deep neural networks, which are trained for individual modalities to learn two deep nonlinear transformations for visual feature extraction from the co-embedding feature space. Second, we propose the global loss function for the DCJN, consisting of a discriminative loss and a correlation loss. The discriminative loss aims to minimize the intraclass distance of the extracted features and maximize the interclass distance of such features to a large margin within each modality, while the correlation loss focuses on mitigating the distribution discrepancy across different modalities. Consequently, the proposed method can realize cross-modality feature extraction guided by the defined global loss function to benefit the similarity measure between 2-D images and 3-D models. For a comparison experiment, we contribute the current largest 2-D image-based 3-D model retrieval dataset. Moreover, the proposed method was further evaluated on three popular benchmarks, including the 3-D Shape Retrieval Contest 2014, 2016, and 2018 benchmarks. The extensive comparison experimental results demonstrate the superiority of this method over the state-of-the-art methods.
Assuntos
Aprendizado Profundo , Diagnóstico por Imagem , Algoritmos , Humanos , Modelos BiológicosRESUMO
Abstract-domain adaptation action recognition is a hot research topic in machine learning and some effective approaches have been proposed. However, samples in the target domain with label information are often required by these approaches. Moreover, domain-invariant discriminative feature learning, feature fusion, and classifier module learning have not been explored in an end-to-end framework. Thus, in this study, we propose a novel end-to-end multiple-view adversarial learning network (MAN) for unsupervised domain adaptation action recognition in which the fusion of RGB and optical-flow features, domain-invariant discrimination feature learning, and action recognition is conducted in a unified framework. Specifically, a robust spatiotemporal feature extraction network, including a spatial transform network and an adaptive intrachannel weight network, is proposed to improve the scale invariance and robustness of the method. Then, a self-attention mechanism fusion module is designed to adaptively fuse the RGB and optical-flow features. Moreover, a multiview adversarial learning loss is developed to obtain domain-invariant discriminative features. In addition, three benchmark datasets are constructed for unsupervised domain adaptation action recognition, for which all actions and samples are carefully collected from public action datasets, and their action categories are hierarchically augmented, which can guide how to extend existing action datasets. We conduct extensive experiments on four benchmark datasets, and the experimental results demonstrate that our proposed MAN can outperform several state-of-the-art unsupervised domain adaptation action recognition approaches. When the SDAI Action II-6 and SDAI Action II-11 datasets are used, MAN can achieve 3.7% ( H â U ) and 6.1% ( H â U ) improvements over the temporal attentive adversarial adaptation network (published in ICCV 2019) module, respectively. As an added contribution, the SDAI Action II-6, SDAI Action II-11, and SDAI Action II-16 datasets will be released to facilitate future research on domain adaptation action recognition.
Assuntos
Aprendizado de Máquina , HumanosRESUMO
Monocular image-based 3-D model retrieval aims to search for relevant 3-D models from a dataset given one RGB image captured in the real world, which can significantly benefit several applications, such as self-service checkout, online shopping, etc. To help advance this promising yet challenging research topic, we built a novel dataset and organized the first international contest for monocular image-based 3-D model retrieval. Moreover, we conduct a thorough analysis of the state-of-the-art methods. Existing methods can be classified into supervised and unsupervised methods. The supervised methods can be analyzed based on several important aspects, such as the strategies of domain adaptation, view fusion, loss function, and similarity measure. The unsupervised methods focus on solving this problem with unlabeled data and domain adaptation. Seven popular metrics are employed to evaluate the performance, and accordingly, we provide a thorough analysis and guidance for future work. To the best of our knowledge, this is the first benchmark for monocular image-based 3-D model retrieval, which aims to help related research in multiview feature learning, domain adaptation, and information retrieval.
Assuntos
Algoritmos , Benchmarking , Armazenamento e Recuperação da InformaçãoRESUMO
Scene graph generation (SGGen) is a challenging task due to a complex visual context of an image. Intuitively, the human visual system can volitionally focus on attended regions by salient stimuli associated with visual cues. For example, to infer the relationship between man and horse, the interaction between human leg and horseback can provide strong visual evidence to predict the predicate ride. Besides, the attended region face can also help to determine the object man. Till now, most of the existing works studied the SGGen by extracting coarse-grained bounding box features while understanding fine-grained visual regions received limited attention. To mitigate the drawback, this article proposes a region-aware attention learning method. The key idea is to explicitly construct the attention space to explore salient regions with the object and predicate inferences. First, we extract a set of regions in an image with the standard detection pipeline. Each region regresses to an object. Second, we propose the object-wise attention graph neural network (GNN), which incorporates attention modules into the graph structure to discover attended regions for object inference. Third, we build the predicate-wise co-attention GNN to jointly highlight subject's and object's attended regions for predicate inference. Particularly, each subject-object pair is connected with one of the latent predicates to construct one triplet. The proposed intra-triplet and inter-triplet learning mechanism can help discover the pair-wise attended regions to infer predicates. Extensive experiments on two popular benchmarks demonstrate the superiority of the proposed method. Additional ablation studies and visualization further validate its effectiveness.
Assuntos
Atenção , Redes Neurais de Computação , Masculino , Humanos , Cavalos , Animais , AprendizagemRESUMO
In this work, we target cross-domain action recognition (CDAR) in the video domain and propose a novel end-to-end pairwise two-stream ConvNets (PTC) algorithm for real-life conditions, in which only a few labeled samples are available. To cope with the limited training sample problem, we employ pairwise network architecture that can leverage training samples from a source domain and, thus, requires only a few labeled samples per category from the target domain. In particular, a frame self-attention mechanism and an adaptive weight scheme are embedded into the PTC network to adaptively combine the RGB and flow features. This design can effectively learn domain-invariant features for both the source and target domains. In addition, we propose a sphere boundary sample-selecting scheme that selects the training samples at the boundary of a class (in the feature space) to train the PTC model. In this way, a well-enhanced generalization capability can be achieved. To validate the effectiveness of our PTC model, we construct two CDAR data sets (SDAI Action I and SDAI Action II) that include indoor and outdoor environments; all actions and samples in these data sets were carefully collected from public action data sets. To the best of our knowledge, these are the first data sets specifically designed for the CDAR task. Extensive experiments were conducted on these two data sets. The results show that PTC outperforms state-of-the-art video action recognition methods in terms of both accuracy and training efficiency. It is noteworthy that when only two labeled training samples per category are used in the SDAI Action I data set, PTC achieves 21.9% and 6.8% improvement in accuracy over two-stream and temporal segment networks models, respectively. As an added contribution, the SDAI Action I and SDAI Action II data sets will be released to facilitate future research on the CDAR task.