Human action recognition in videos with articulated pose information by deep networks

Research on the Recognition Method of the Axle End Mark of a Train Wheelset Based on Machine Vision

Whether the wheelset of a high-speed train has defects such as cracks is very important to the safety of high-speed trains. Hence, the wheelset must be regularly inspected for flaws. For flaw detection of a wheelset, it is necessary to record the axle end information of the wheelset to correlate with the flaw detection results. To quickly and accurately identify the axle end mark of the wheelset, an automatic identification method based on machine vision is proposed. Our method identifies seven types of marks on the axle end, including the smelting number, steel grade number, unit number, sequence number, year and month, axle type mark, and the azimuth mark. Using the established automatic identification method of axle end marks, based on Retinex theory, an improved dual-core Laplacian combined with Gaussian filtering operation is proposed to solve the problem of the low contrast of the wheelset axle end image. An improved image tilt correction algorithm based on the combination of Hough circle detection and bilinear interpolation is proposed, which solves the angle tilt problem of the target character area of the axis end image. To handle the various types of axis end markers and the small amount of data, a retraining method to improve recognition accuracy is proposed. This method first uses Chi_Sim as the basic font for training and then retrains based on the trained font. Finally, Tesseract-OCR is used to improve the accuracy of the recognition results. Experiments are carried out by developing an automatic recognition program for axle end marks. The results show that the proposed method can effectively identify and classify seven-character types, and the recognition accuracy reaches 96.88% while the recognition time of each image is 5.88 s.

Attention mechanism and mixup data augmentation for classification of COVID-19 Computed Tomography images

The Coronavirus disease is quickly spreading all over the world and the emergency situation is still out of control. Latest achievements of deep learning algorithms suggest the use of deep Convolutional Neural Network to implement a computer-aided diagnostic system for automatic classification of COVID-19 CT images. In this paper, we propose to employ a feature-wise attention layer in order to enhance the discriminative features obtained by convolutional networks. Moreover, the original performance of the network has been improved using the mixup data augmentation technique. This work compares the proposed attention-based model against the stacked attention networks, and traditional versus mixup data augmentation approaches. We deduced that feature-wise attention extension, while outperforming the stacked attention variants, achieves remarkable improvements over the baseline convolutional neural networks. That is, ResNet50 architecture extended with a feature-wise attention layer obtained 95.57% accuracy score, which, to best of our knowledge, fixes the state-of-the-art in the challenging COVID-CT dataset.