Search

Diabetic retinopathy (DR) is a serious ocular complication that can pose a serious risk to a patient's vision and overall health. Currently, the automatic grading of DR is mainly using deep learning techniques. However, the lesion information in DR images is complex, variable in shape and size, and randomly distributed in the images, which leads to some shortcomings of the current research methods, i.e., it is difficult to effectively extract the information of these various features, and it is difficult to establish the connection between the lesion information in different regions. To address these shortcomings, we design a multi-scale dynamic fusion (MSDF) module and combine it with graph convolution operations to propose a multi-scale dynamic graph convolutional network (MDGNet) in this paper. MDGNet firstly uses convolution kernels with different sizes to extract features with different shapes and sizes in the lesion regions, and then automatically learns the corresponding weights for feature fusion according to the contribution of different features to model grading. Finally, the graph convolution operation is used to link the lesion features in different regions. As a result, our proposed method can effectively combine local and global features, which is beneficial for the correct DR grading. We evaluate the effectiveness of method on two publicly available datasets, namely APTOS and DDR. Extensive experiments demonstrate that our proposed MDGNet achieves the best grading results on APTOS and DDR, and is more accurate and diverse for the extraction of lesion information. [ABSTRACT FROM AUTHOR]

In this research, Lithium-ion (Li-ion) batteries were tested at four different charge rates (DCR): 0.2C, 0.5C, 1C, and 1.5C, and four different discharge rates (DDR): 0.5C, 0.9C, 1.3C, 1.6C. This paper proposes a capacity fade model for charging and discharging at accelerated current-rate (C-rate), to interpret the vulnerabilities of Li-ion batteries in energy storage system, because Lithium-ion (Li-ion) batteries are prone to ageing at the fluctuation of the loads in micro-grids. The characteristics of Li-ion batteries both at accelerated DCR and DDR are thoroughly investigated. It is discovered that charging and discharging Li-ion batteries outside of the standard C-rate accelerates their ageing. In addition, the degree of capacity fade is assessed at an accelerated C-rate to develop an ideal charge and discharge model for the micro-grids. Furthermore, the battery capacity fade model is then investigated with deep learning algorithm-based feed-forward neural network (FNN), and recurrent neural network with long-short term memory layer (RNN-LSTM). A comparison of the developed capacity fade models is performed, and it is discovered that the LSTM-RNN battery ageing model outperforms the conventional FNN network at accelerated C-rate. Nevertheless, the error metrics performance of both FNN and LSTM-RNN are less than 0.1%. [ABSTRACT FROM AUTHOR]

Video retrieval methods, e.g., for visual concept classification, person recognition, and similarity search, are essential to perform fine-grained semantic search in large video archives. However, such retrieval methods often have to be adapted to the users' changing search requirements: which concepts or persons are frequently searched for, what research topics are currently important or will be relevant in the future? In this paper, we present VIVA, a software tool for building content-based video retrieval methods based on deep learning models. VIVA allows non-expert users to conduct visual information retrieval for concepts and persons in video archives and to add new people or concepts to the underlying deep learning models as new requirements arise. For this purpose, VIVA provides a novel semi-automatic data acquisition workflow including a web crawler, image similarity search, as well as review and user feedback components to reduce the time-consuming manual effort for collecting training samples. We present experimental retrieval results using VIVA for four use cases in the context of a historical video collection of the German Broadcasting Archive based on about 34,000 h of television recordings from the former German Democratic Republic (GDR). We evaluate the performance of deep learning models built using VIVA for 91 GDR specific concepts and 98 personalities from the former GDR as well as the performance of the image and person similarity search approaches. [ABSTRACT FROM AUTHOR]

The German Broadcasting Archive maintains the cultural heritage of radio and television broadcasts of the former German Democratic Republic (GDR). The uniqueness and importance of the video material fosters a large scientific interest in the video content. In this paper, we present a system for automatic video content analysis and retrieval to facilitate search in historical collections of GDR television recordings. It relies on a distributed, service-oriented architecture and includes video analysis algorithms for shot boundary detection, concept classification, person recognition, text recognition and similarity search. The combination of different search modalities allows users to obtain answers for a wide range of queries, leading to satisfactory results in short time. The performance of the system is evaluated using 2500 h of GDR television recordings. [ABSTRACT FROM AUTHOR]