Precise and reliable forecasting of short-term electricity load is essential to the development of smart grids. Particularly, deep neural networks (DNNs) are widely utilized for the prediction of shortterm electricity load due to their automatic feature extraction ability. However, these available stacked deeplearning models may lose some temporal features or spatial features of original input data. To capture more comprehensive information, in this paper, we present an integration scheme based on empirical mode decomposition (EMD), similar day methods, and DNNs to perform short-term load forecasting. It is especially worth noting that the electricity price is also an important factor for load variation, which is considered in our proposed scheme. Specifically, there are two primary layers: a feature extraction layer and a forecasting layer. In the feature extraction layer, EMD is applied to decompose load time series into several components, which are arranged into the 2-D input matrix of the convolutional neural network (CNN). Both the output vectors of the CNN and the raw load sequences are fed into the long short-term memory (LSTM) layer. Therefore, the whole EMD based CNN-LSTM approach extracts multimodal spatial-temporal features from input data. Meanwhile, the electricity price data is utilized to obtain multimodal spatial-temporal features in the same way. Additionally, the day and hour information and loads of similar days are to augment extra features for prediction. In the forecasting layer, the forecasting task is accomplished through a fully-connected neural network based on the outputs of the feature extraction layer. Leveraging these techniques enables our proposed scheme to extract more latent features, which significantly improve the accuracy. In order to demonstrate the performance of our proposed scheme, related experiments are conducted on actual data from the electricity market in Singapore. Compared to other available models, our proposed scheme is superior in graphic and numerical results. INDEX TERMS Short-term load forecasting, empirical mode decomposition, similar day methods, deep neural networks, transitional forecasting scheme, electricity price, multimodal spatial-temporal features.

The panoramic dental X-ray images are an essential diagnostic tool used by dentists to detect the symptoms in an early stage and develop appropriate treatment plans. In recent years, deep learning methods have been applied to achieve tooth segmentation of dental X-rays, which aims to assist dentists in making clinical decisions. Because the original images contain plenty of useless information, it is necessary to extract the region-of-interest (ROI) to obtain more accurate results by focusing on the maxillofacial region. However, a fast and accurate maxillofacial segmentation without hand-crafted features is challenging due to the poor image quality. In this study, we create a large maxillofacial dataset and propose an efficient encoder-decoder network model named EED-Net to solve this problem. This dataset consists of 2602 panoramic dental X-ray images and corresponding segmentation masks annotated by the trained experts. Based on the original structure of U-Net, our model structure contains three major modules: a feature encoder, a corresponding decoder, and a multipath feature extractor that connects the encoding path and the decoding path. In order to obtain more semantic features from the depth and breadth, we replace the convolution layer with the residual block in the encoder and adopt Inception-ResNet block in the multipath feature extractor. Inspired by the skip connection in FCN-8s, the lightweight decoder has the same channel dimension as the number of segmented objects. Besides, a weighted loss function is used to enhance segmentation accuracy. The comprehensive experimental results on the new dataset demonstrate that our model achieves better accuracy and speed trade-offs for maxillofacial segmentation than the latest methods.

In recent years, few-shot learning is proposed to solve the problem of lacking samples in deep learning. However, previous works are mainly concentrated on optimizing neural network structures or augmenting the dataset while ignoring the local relationship of the images. Considering that humans pay more attention to the foreground or prominent features of the images during image recognition, we proposed the foreground feature attention module (FFAM) based on an unsupervised saliency detector for few-shot learning. The FFAM consists of two parts: the foreground extraction module and the features attention module. More specifically, we first extract the foreground images by Robust Background Detector (RBD), one of the best unsupervised saliency detectors. Secondly, we employ the same embedding module to extract the features of both original images and foreground images. Finally, we introduce three improvements to enhance the foreground features and make our network focus on the foreground features without losing background information. Our proposed FFAM is more sensitive to the foreground features than previous approaches. Hence, it effectively recognizes those images with similar backgrounds. Extensive experiments are conducted on miniImagenet and tieredImagenet datasets. It is demonstrated that our proposed FFAM greatly improves the accuracy performance over baseline systems for both one-shot and few-shot classification tasks without increasing the network complexity.