As an important part of intelligent surveillance systems, person re-identification (PReID) has drawn wide attention of the public in recent years. Many recent deep learning-based PReID methods have used attention or multi-scale feature learning modules to enhance the discrimination of the learned deep features. However, the attention mechanisms may lose some important feature information. Moreover, the multi-scale models usually embed the multi-scale feature learning module into the backbone network, which increases the complexity of testing network. To address the two issues, we propose a multi-scale deep supervision with attention feature learning deep model for PReID. Specifically, we introduce a reverse attention module to remedy the feature information losing issue caused by the attention module, and a multi-scale feature learning layer with deep supervision to train the network. The proposed modules are only used at the training phase and discarded during the test phase. Experiments on Market-1501, DukeMTMC-reID, CUHK03 and MSMT17 datasets. demonstrate that our model notably beats other competitive state-of-the-art models.

Person re-identification (PReID) has received increasing attention due to its significantly importance in intelligent video surveillance. However, most existing multi-scale feature learning methods embed the multi-scale feature extraction modules for PReID, which increases the complexity of inference network and reduces the timeliness. Moreover, jointly using the small-scale and large-scale features to learn feature representations may weaken the local detailed features extraction and spatial information learning. Besides, some attentive local features are often suppressed when introducing the attention mechanisms for deep PReID models. To address these issues, a deep model with Multi-scale Feature Representation Learning (MFRL) and Random Batch Feature Mask (RBFM) is proposed for PReID in this study. To ensure the feature representations discriminability and spatial information learning, two identity losses are adopted to supervise the small-scale and large-scale features learning in MFRL module, respectively. To alleviate the situation of local attentive features being suppressed by using attention mechanisms, RBFM branch with random feature block dropping strategy which can learn the attentive local feature representations. The proposed methods are only performed in the training phase and discarded in the testing phase, thus, enhancing the effectiveness of the model. Our model achieves the state-of-the-art on the popular benchmark datasets including Market-1501, DukeMTMC-reID and CUHK03. Besides, we conduct a set of ablation experiments to verify the effectiveness of the proposed methods.

Changing of brain resulting from brainfag is discussed in this paper. According to the system cluster analysis on EEG data of the subjects, we can get the methods of estimating brain fatigue. Moreover, this study also provides us the theoretical basis and some basic experimental evidence to make a scientific forecast of brainfag. Twelve right-handed, no-smoking healthy subjects took part in the study. There are many events that can trigger the brainfag, so we did different experiments that included closing eyes, ingesting some alcohol, counting and hearing. EEG data were recorded with the international 10/20 system, and then we use clustering method to analyze the EEG data. The method of analysis is system cluster analysis analysis. Processing and statistic analysis of EEG data has performed using EEG analysis toolbox and program of system cluster analysis. After EEG signal analysis, we found that the electrodes were separated into three clusters located in frontal & central area, occipital area and temporal area in the normal state. Most electrodes in frontal, central and occipital area got together and formed into a cluster after they ingested some alcohol, and then the cluster separated gradually with the increase of ingesting alcohol. After counting, the EEG signal from left-brain was very different from signals from right-brain and the clusters mostly were distributed longitudinally. After hearing, the EEG signal from left-brain was also different from signals from right-brain and the result of clustering was the same as counting. All the phenomenons tell us that A little alcohol can strong the relatives of subjectspsila frontal and occipital head EEG signal. At the same time, due to alpha rhythm is a stronger part between 4 Hz and 30 Hz, alpha rhythm is gradually enhanced and the area of alpha rhythm is expanded after alcohol ingestion, however, the cluster separated gradually with the increase of ingesting alcohol. This indicate that too much drinking maybe lead to the decrease of equilibrium, stability between the parts of brain and the increase of containment procedure and brainfag. The left brain is in charge of counting and comprehension of language, so the left brain was in the state of fatigue during counting and hearing. The change led to the correlation weakening between brainpsilas left area and right area and then longitudinal distribution of the clusters appeared.

So far, slope collapse detection mainly depends on manpower, which has the following drawbacks: (1) low reliability, (2) high risk of human safe, (3) high labor cost. To improve the efficiency and reduce the human investment of slope collapse detection, this paper proposes an intelligent detection method based on deep learning technology for the task. In this method, we first use the deep learning-based image segmentation technology to find the slope area from the captured scene image. Then the foreground motion detection method is used for detecting the motion of the slope area. Finally, we design a lightweight convolutional neural network with an attention mechanism to recognize the detected motion object, thus eliminating the interference motion and increasing the detection accuracy rate. Experimental results on the artificial data and relevant scene data show that the proposed detection method can effectively identify the slope collapse, which has its applicative value and brilliant prospect.

Sentence similarity is a fundamental task to measure the degree of the likelihood between two sentences. It plays an important role in both NLP and IR communities. The current sentence semantic similarity measurements can be divided into two categories: the model based on feature engineering and the model based on deep learning. These two categories all have their own advantages and disadvantages. The feature captured based on deep learning may not get the deep meaning of the sentence. And the feature exacted based on feature engineering may not be comprehensive. The algorithm proposed recently usually use one of these two models. In this paper, we proposed a new model that combine the deep learning model and the feature engineering model together. Our model enables two methods to complement each other, making the feature extraction more comprehensive: not only take the global semantic information into account, but also further refine the semantic information at the word level. The experiment shows that our model performs better than models only use one category.