Vehicle detection based on very high-resolution (VHR) remote sensing images is beneficial in many fields such as military surveillance, traffic control, and social/economic studies. However, intricate details about the vehicle and the surrounding background provided by VHR images require sophisticated analysis based on massive data samples, though the number of reliable labeled training data is limited. In practice, data augmentation is often leveraged to solve this conflict. The traditional data augmentation strategy uses a combination of rotation, scaling, and flipping transformations, etc., and has limited capabilities in capturing the essence of feature distribution and proving data diversity. In this study, we propose a learning method named Vehicle Synthesis Generative Adversarial Networks (VS-GANs) to generate annotated vehicles from remote sensing images. The proposed framework has one generator and two discriminators, which try to synthesize realistic vehicles and learn the background context simultaneously. The method can quickly generate high-quality annotated vehicle data samples and greatly helps in the training of vehicle detectors. Experimental results show that the proposed framework can synthesize vehicles and their background images with variations and different levels of details. Compared with traditional data augmentation methods, the proposed method significantly improves the generalization capability of vehicle detectors. Finally, the contribution of VS-GANs to vehicle detection in VHR remote sensing images was proved in experiments conducted on UCAS-AOD and NWPU VHR-10 datasets using up-to-date target detection frameworks.

The evaluations of traditional teaching quality are mainly subjective, and there is a lack of fine-grained objective data to support the evaluation of teaching states in the classroom. In this paper, an intensity-based facial expression dataset is proposed and named EIDB-13, which contains 13 kinds and 10393 facial images collected from thousands of individuals and existing facial expression datasets. Convolutional neural network (CNN) and attention mechanism are combined to recognize facial expressions. Migration learning is used to solve over-fitting problem in the process of training deep network based on the small sample dataset. InceptionResNetV2 is employed as migration network. Furthermore, an InceptionResNetV2+CBAM network proposed extract similar feature information among facial expressions and it outperforms the network without attention mechanisms. Experiments show a classification accuracy rate of 78% on the intensity-based facial expression dataset EIDB-13 and of 88% on the public macro expression dataset RAF-DB. Combining facial expression recognition technology into teaching is a key foundation to study teaching quality on the intensity of teacher’s expression.

Emotion recognition from macroexpression and microexpression has been widely used in applications such as human–computer interaction, learning status evaluation, and mental disorder diagnosis. However, due to the complexity of human macroexpressions, recognizing macroexpressions with high accuracy is a challenging task. Moreover, the short duration and low movement intensity of microexpressions make its recognition more difficult. For MM-FER (macro and microfacial expression recognition), the key information can be more efficiently expressed by a graph. In this article, a novel framework based on graph neural network named SSGNN (spatial and spectral domain features based on a graph neural network) is designed to extract spatial and spectral domain features from facial images for MM-FER, which can efficiently recognize both macroexpressions and microexpressions under the same model. SSGNN consists of two parts, SPAGNN and SPEGNN, which are used to extract spectral and spatial domain features, respectively. Experiments proved that jointly using the spectral and spatial information extracted by SSGNN can largely improve the performance of MM-FER when the training sample is limited. First, the influences of different neighbors and samples to the model performance was analyzed. Then, the contribution of SPAGNN and SPEGNN were evaluated. It was discovered that fusing the result of SPAGNN and SPEGNN at decision level further improved the performance of MM-FER. Experiment proved that SSGNN can recognize microexpression acquired by various sensors with higher accuracy under different image resolutions and image formats than the compared state-of-the-art methods in most cases. A cross-dataset experiment demonstrated the generalization ability of SSGNN.

Research shows that physiological signals can provide objective data support for the analysis of human emotions. At present, non-contact heart rate data have been employed in the research of medicine, intelligent transportation, smart education, etc. However, it is hard to detect heart rate data using non-contact traditional methods during head rotation, especially when face information is missing in scenarios such as online teaching/learning. Traditional remote photoplethysmography (rPPG) methods require a static, full frontal face within a fixed distance for heart rate detection. These strict requirements make it impractical to measure heart rate data in real-world scenarios, as a lot of videos only partially record the subjects’ face information, such as profile, too small distance, and wearing a mask. The current algorithm aims to solve the problem of head deflections between 30 degrees and 45 degrees by employing a symmetry substitution method, which can replace the undetected region of interest (ROI) with the detectable one. When face information is partially missing, our algorithm uses face–eye location to determine ROI. The results show that the method in this paper can solve certain practical problems related to heart rate detection, with a root mean square error (RMSE) under 7.64 bpm.