Zero-shot learning (ZSL) is made possible by a projection function between feature space and semantic (e.g., an attribute space). Key to ZSL thus learn that robust against the often large domain gap seen unseen class domains. In this work, achieved data synthesis learning. Specifically, novel strategy proposed, which prototypes vectors) are used simply perturb for generating ones. As in any syn...

Analyzing land cover using remote sensing images has broad prospects, the precise segmentation of is key to application this technology. Nowadays, Convolution Neural Network (CNN) widely used in many image semantic tasks. However, existing CNN models often exhibit poor generalization ability and low accuracy when dealing with To solve problem, paper proposes Dual Function Feature Aggregation (D...

While deep learning demonstrates its strong ability to handle independent and identically distributed (IID) data, it often suffers from out-of-distribution (OoD) generalization, where the test data come another distribution (w.r.t. training one). Designing a general OoD generalization framework for wide range of applications is challenging, mainly due different kinds shifts in real world, such ...

This paper shows that it is possible to build a system with Semantic Desktop functionality — the annotation and linking of desktop data across application boundaries, and the usage of such links to achieve new ways of organizing and finding data in a meaningful way — in a very simple and light-weight fashion. Instead of building an extensive and feature-rich system from scratch, we harness the ...

We introduce new features for incorporating semantic predicate-argument structures in machine translation (MT). The methods focus on the completeness of the semantic structures of the translations, as well as the order of the translated semantic roles. We experiment with translation rules which contain the core arguments for the predicates in the source side of a MT system, and observe that usi...

Our feature configuration tool ST 2 Configurator integrates (1) a visual interactive representation of the feature model and (2) a formal reasoning engine that calculates consequences of the user’s actions and provides formal explanations. The tool’s software architecture is designed as a chain of components, which provide mappings between visual elements and their corresponding formal represen...

Geospatial feature discovery from remote sensing imageries is widely used in national defense and security communities. Existing methods in the geospatial image mining and feature extraction focus on the manual or automated processing of images to detect individual elementary features, such as building and highway. Such elementary features don’t tell much semantic information about the features...