Diabetic retinopathy (DR) is an important causes of blindness worldwide. However, DR is hard to detected in early stages and the diagnostic procedure can be time-consuming even for experienced experts. Therefore, a computer-aided diagnosis method based on deep learning algorithms is proposed to automatedly diagnose the referable deiabetic retinopathy (RDR) by classifying color retinal fundus photographs into two grades. In this work, A novel convolutional neural network model with Siamese-like architecture is trained with transfer learning technique. Different from previous works, the proposed model accepts binocular fundus images as inputs and learns their correlation to help making prediction. In the case with a training set of only 28104 images and a test set of 7024 images, an area under the receiver operating curve (AUC) of 0.951 is obtained by the proposed binocular model, which is 0.011 higher than that obtained by existing monocular model. To further verify the effectiveness of the binocular design, a binocular model for five-class DR detection is also trained and evaluated on a 10% validation set. The result shows that it achieves a kappa score of 0.829 which is higher than that of existing non-emsemble model.

In this work, we propose a novel Deep Convolutional Neural Network (DCNN) tailored for gas classification. Inspired by the great success of DCNN in the field of computer vision, we designed a DCNN with up to 38 layers. In general, the proposed gas neural network, named GasNet, consists of: six convolutional blocks, each block consist of six layers; a pooling layer; and a fully-connected layer. Together, these various layers make up a powerful deep model for gas classification. Experimental results show that the proposed DCNN method is an effective technique for classifying electronic nose data. We also demonstrate that the DCNN method can provide higher classification accuracy than comparable Support Vector Machine (SVM) methods and Multiple Layer Perceptron (MLP).

Many deep learning (DL) models have shown exceptional promise in radar-based human activity recognition (HAR) area. For radar-based HAR, the raw data is generally converted into a 2-D spectrogram by using short-time Fourier transform (STFT). All the existing DL methods treat the spectrogram as an optical image, and thus the corresponding architectures such as 2-D convolutional neural networks (2D-CNNs) are adopted in those methods. These 2-D methods that ignore temporal characteristics ordinarily lead to a complex network with a huge amount of parameters but limited recognition accuracy. In this paper, for the first time, the radar spectrogram is treated as a time sequence with multiple channels. Hence, we propose a DL model composed of 1-D convolutional neural networks (1D-CNNs) and long short-term memory (LSTM). The experiments results show that the proposed model can extract spatio-temporal characteristics of the radar data and thus achieves the best recognition accuracy and relatively low complexity compared to the existing 2D-CNN methods.

Radar-based human activity recognition (HAR) has great potential in many fields, such as surveillance, smart homes, and human-computer interaction. Complex deep neural networks have brought significant improvement in classification performance but also a surge of computational cost and the number of parameters, which makes it challenging to deploy in mobile devices. However, the existing studies in this area mainly focus on improving the classification accuracy. In this article, we propose an extremely efficient convolutional neural network (CNN) architecture named Mobile-RadarNet, which is specially designed for human activity classification based on micro-Doppler signatures. The new architecture exploits 1-D depthwise convolutions and pointwise convolutions to build lightweight CNN architecture. The experiments on a seven-class human activity data set demonstrate that the proposed Mobile-RadarNet can achieve high classification accuracy meanwhile to keep the computational complexity at an extremely low level, and thus has great potential to be deployed in the mobile devices.

In this paper, we present a novel one-dimensional deep convolutional neural network (1D-DCNN) with a multilabel way based algorithm for comprehensively and automatically extracting features and classifying mixture gases. Although a number of pattern recognition methods have been used to analyze the mixed gases, the performances of these methods highly depend on the hand-crafted feature engineering. By contrast, the proposed implementation, based on one-dimensional convolution, is capable of automatically extracting features and distinguishing the individual component of binary mixture gases composed of Ethylene, CO and Methane. To the best of our knowledge, the proposed 1D-DCNN algorithm is firstly applied in the mixture gases recognition. In addition, the proposed 1D-DCNN with multilabel way not only significantly reduces the label dimension but also quantifies the probability of each component in mixed gases. Compared with the conventional pattern recognition algorithms including support vector machine (SVM), artificial neural network (ANN), k-nearest neighbor (KNN) and random forest (RF), the proposed 1D-DCNN exhibits higher recognition accuracy (96.30%) based on our extensive experimental results using 10- fold cross validation.

Division of focal plane (DoFP) polarimeter is widely used in polarization imaging sensors. The periodically arranged micro-polarizers integrated on the focal plane ensure its outstanding real-time performance, but reduce the spatial resolution of output images and further affect the calculation of polarization parameters. In this paper, a four-layer, end-to-end fully convolutional neural network called Fork-Net is proposed, which aims to directly improve the imaging quality of three polarization properties: intensity (i.e., S₀), degree of linear polarization (DoLP), and angle of polarization (AoP), rather than focusing on reducing the interpolation error of intensity images of different polarization orientations. The Fork-Net accepts raw mosaic images as input and directly outputs S₀, DoLP, and AoP. It is also trained with a customized loss function. The experimental results show that compared with existing methods, the proposed one achieves the highest peak signal-to-noise ratio (PSNR) and prominent visual quality on output images.