Image fusion is an important task for computer vision as a diverse range of applications are benefiting from the fusion operation. The existing image fusion methods are largely implemented at the pixel level, which ma...
详细信息
Image fusion is an important task for computer vision as a diverse range of applications are benefiting from the fusion operation. The existing image fusion methods are largely implemented at the pixel level, which may introduce artifacts and/or inconsistencies, while the computational complexity is relatively high. In this article, we propose a symmetric encoder-decoder with residual block (SEDRFuse) network to fuse infrared and visible images for night vision applications. At the training stage, the SEDRFuse network is trained to create a fixed feature extractor. At the fusing stage, the trained extractor is utilized to extract the intermediate and compensation features, which are generated by the residual block and the first two convolutional layers from the input source images, respectively. Two attention maps, which are derived from the intermediate features, are then multiplied by the intermediate features for fusion. The salient compensation features obtained through elementwise selection are passed to the corresponding deconvolutional layers for processing. Finally, the fused intermediate features and the selected compensation features are decoded to reconstruct the fused image. Experimental results demonstrate that the proposed fusion solution, i.e., SEDRFuse, outperforms the state-of-the-art fusion methods in terms of both subjective and objective evaluations.
Traditionally, the training phase of abstractive text summarization involves inputting two sets of integer sequences;the first set representing the source text, and the second set representing words existing in the re...
详细信息
ISBN:
(数字)9781665457279
ISBN:
(纸本)9781665457279
Traditionally, the training phase of abstractive text summarization involves inputting two sets of integer sequences;the first set representing the source text, and the second set representing words existing in the reference summary, into the encoder and decoder parts of the model, respectively. However, by using this method, the model tends to perform poorly if the source text includes words which are irrelevant or insignificant to the key ideas. In order to address this issue, we propose a new keywords-based method for abstractive summarization by combining the information provided by the source text and its keywords to generate summary. We utilize a bi-directional long short-term memory model for keyword labelling, using overlapping words between the source text and the reference summary as ground truth. The results obtained from our experiments on ThaiSum dataset show that our proposed method outperforms the traditional encoder-decoder model by 0.0425 on ROUGE-1 F1, 0.0301 on ROUGE-2 F1 and 0.0140 on BERTScore F1.
Ship speed prediction is the basis for the realization of ship intelligence which can reduce energy consumption and protect the environment. We aim to propose a model for predicting accurate and timely ship speed. The...
详细信息
ISBN:
(数字)9781665482905
ISBN:
(纸本)9781665482905
Ship speed prediction is the basis for the realization of ship intelligence which can reduce energy consumption and protect the environment. We aim to propose a model for predicting accurate and timely ship speed. The speed prediction belongs to time series forecasting. In order to leverage the previous data among multi horizons and find highly non- linear relationships in long time range, we Introduce a Gated Recurrent Unit (GRU) based encoderdecoder with temporal attention mechanism for speed prediction. The attention mechanism is adopted to assign different weights to ship speed of the input time steps. To validate the effectiveness of our model, we choose three baseline models to train and test on the same ship navigation dataset. The comparative experiment results suggest that our model has lower RMSE and MAE than the others. The proposed model in our paper performs better in ship speed prediction.
With the rapid development of Internet technology and social media, people are accustomed to making comments on the Internet. Sentiment analysis, as an efficient technique, has been used by researchers in the tasks of...
详细信息
ISBN:
(纸本)9781728140346
With the rapid development of Internet technology and social media, people are accustomed to making comments on the Internet. Sentiment analysis, as an efficient technique, has been used by researchers in the tasks of analysing the sentiment polarity under these comments. To better achieve this target, the fundamental challenge is how to extract the feature and build a proper mechanism to learn them. A lot of word embedding based deep learning models for sentiment analysis are proposed in the literature. And the semi-supervised learning methods make it possible to use both labelled and unlabelled data for this kind of task. Furthermore, the attention mechanism proposed in recent years has achieved great accomplishments for natural language processing (NLP) tasks since it helps to capture the important information of the documents. In this paper, inspired by these works, we proposed a long short term memory (LSTM) based semi-supervised attention framework for sentiment analysis tasks, which is composed of an unsupervised attention based LSTM encoder-decoder and an attention based supervised LSTM model attached by a Softmax layer. The unsupervised part worked for attaining the high dimensional representation of the documents, and the supervised part extracted feature and enhanced the important parts for classification. Experimental study on commonly used datasets has demonstrated its ability for sentiment analysis tasks.
暂无评论