Search

Objects in a real scene often occlude each other and inferring a complete appearance from the visible part is an important and challenging task. In this paper, the authors propose a self‐supervised generative adversarial network GIGAN (GAN for generating the invisible), which can generate the complete appearance of objects without labelled invisible part information. The authors build two cycle transformation networks CycleIncomplete (CycleI) and CycleComplete (CycleC) that share parameters to improve the accuracy of mask completion. This design does not require well‐matched training images and can make better use of the limited labelled samples. In addition, the authors propose a conditional normalization module and combine it with the inferred complete mask output. The combination not only enhances the content recovery ability and obtains more realistic outputs, but also improves the efficiency of the generation process. Experimental results show that compared with existing self‐supervised learning models, our method achieves l1 error, mean intersection‐over‐union (mIOU), and Fréchet inception distance (FID) improvements on the COCOA and KINS datasets. [ABSTRACT FROM AUTHOR]

• A multidimensional data secure aggregation scheme in smart grid is proposed. • The scheme achieves fine-grained access control at dimension-level. • Access permission can be updated flexibly and securely. • The proposed scheme is implemented efficiently. In smart grid, for saving communication bandwidth and performing fine-grained analysis, the power data need to be collected and aggregated in multi-dimensions. As each dimension contains sensitive information of users and may be used by different analytic organizations, privacy preserving and access control for multidimensional power data are obviously important. Existing schemes either cannot support dimensional access control or require transmitting attribute sets of data together with ciphertexts, which leaks the relevant information of power data and increases the security risk. Moreover, to disseminate and implement access control of the multidimensional data, control center (CC) must firstly decrypt the data of all dimensions, which may result in the data leakage. In this paper, we propose a Multidimensional-data Tight Aggregation scheme which supports Privacy preserving and fine-grained Access control (MTA–PA). In MTA–PA, we import the extended Boneh–Goh–Nissim (EBGN) homomorphic encryption from wireless sensor networks to aggregate multidimensional power data into one ciphertext and simultaneously use key-policy attribute-based encryption (KP–ABE) to enforce fine-grained access control at the dimension level. By using MTA–PA, the gateway need not to distinguish different ciphertexts before aggregating them, and organizations only can analyze the authorized dimensions of power data. MTA–PA further achieves secure access-permission update via updating related parameters of EBGN and KP–ABE. And then, we also present the implementation methods of the algorithms. The security analysis and experimental results show that MTA–PA is a secure and efficient data protection scheme in smart grid. [ABSTRACT FROM AUTHOR]

This paper proposes a novel Hough-based object shape representation model called Pair Hough Model (PHM) and its corresponding object detection framework. PHM constructs the voting models implicitly with automatically detected interest points and their local descriptors for unseen object categories. In addition, by casting votes according to key point pairs instead of individual key points and taking the orientations of objects as well as their sizes into consideration, PHM can recognize and localize objects after their scaling and/or rotation, which makes it suitable for processing images with major rotations such as pictures taken by mobile devices. Evaluation experiments proved that PHM does not need to be trained on rotated images to recognize rotated objects, and PHM achieved comparable results to the state-of-the-art methods on several widely used public data sets. [ABSTRACT FROM AUTHOR]