Track topics on Twitter Track topics that are important to you
Effectively recognizing different sceneries with complex backgrounds and varied lighting conditions plays an important role in modern AI systems. Competitive performance has recently been achieved by the deep scene categorization models. However, these models implicitly hypothesize that the image-level labels are 100% correct, which is too restrictive. Practically, the image-level labels for massive-scale scenery sets are usually calculated by external predictors such as ImageNet-CN. These labels can easily become contaminated because no predictors are completely accurate. This article proposes a new deep architecture that calculates scene categories by hierarchically deriving stable templates, which are discovered using a generative model. Specifically, we first construct a semantic space by incorporating image-level labels using subspace embedding. Afterward, it is noticeable that in the semantic space, the superpixel distributions from identically labeled images remain unchanged, regardless of the image-level label noises. On the basis of this observation, a probabilistic generative model learns the stable templates for each scene category. To deeply represent each scenery category, a novel aggregation network is developed to statistically concatenate the CNN features learned from scene annotations predicted by HSA. Finally, the learned deep representations are integrated into an image kernel, which is subsequently incorporated into a multiclass SVM for distinguishing scene categories. Thorough experiments have shown the performance of our method. As a byproduct, an empirical study of 33 SIFT-flow categories shows that the learned stable templates remain almost unchanged under a nearly 36% image label contamination rate.
This article was published in the following journal.
Name: IEEE transactions on cybernetics
The serial-position effect in working memory is considered important for studying how a sequence of sensory information can be retained and manipulated simultaneously in neural memory circuits. Here, ...
When new events overlap with past events, there is a natural tradeoff between encoding the new event and retrieving the past event. Given the ubiquity of overlap among memories, this tradeoff between ...
Compound-specific isotope analysis (CSIA) was firstly applied to explore the biodegradation of hexachlorcyclohexane (HCH) isomers in contaminated soil. Concentrations and compound-specific carbon isot...
To explore dietary patterns within the context of the nutrition transition among Mexican adolescents, we employed a mixed-methodology that included survey data from a cohort of 550 adolescents and dir...
Microbial communities play a crucial role in bioremediation of pollutants in contaminated ecosystem. In addition to pure culture isolation and bacterial 16S rRNA based community studies, the focus has...
The project is dedicated to development of new paradigms to investigate memory and attention in the rich and realistic environments. The investigator will use modern interactive digital te...
As part of the ZOSTER-006 and ZOSTER-022 pivotal trials of the HZ/su vaccine, all study participants completed quality of life (QoL) questionnaires. The only questionnaires encoded into th...
Automated Insulin Delivery (AID) systems have now become an important standard-of-care for people with T1D and have demonstrated a reduction, but not elimination, of hypoglycemia during lo...
This multi-center study is to focus on patients with sepsis in Intensive Care Units (ICUs) in order to better understand the complex host-pathogen interaction and clinical heterogeneity as...
This is a prospective, multicenter, interventional, observational, open label, single arm, longitudinal evaluation of ventral incisional hernia repair using LTM in contaminated or infected...
Information application based on a variety of coding methods to minimize the amount of data to be stored, retrieved, or transmitted. Data compression can be applied to various forms of data, such as images and signals. It is used to reduce costs and increase efficiency in the maintenance of large volumes of data.
Various units or machines that operate in combination or in conjunction with a computer but are not physically part of it. Peripheral devices typically display computer data, store data from the computer and return the data to the computer on demand, prepare data for human use, or acquire data from a source and convert it to a form usable by a computer. (Computer Dictionary, 4th ed.)
The science and art of collecting, summarizing, and analyzing data that are subject to random variation. The term is also applied to the data themselves and to the summarization of the data.
Systematic gathering of data for a particular purpose from various sources, including questionnaires, interviews, observation, existing records, and electronic devices. The process is usually preliminary to statistical analysis of the data.
Devices capable of receiving data, retaining data for an indefinite or finite period of time, and supplying data upon demand.
Food is any substance consumed to provide nutritional support for the body. It is usually of plant or animal origin, and contains essential nutrients, such as carbohydrates, fats, proteins, vitamins, or minerals. The substance is ingested by an organism ...