Image classification has always been a hot research direction in the world, and the emergence of deep learning has promoted the development of this field. Convolutional neural networks (CNNs) have gradually become the mainstream algorithm for image classification since 2012, and the CNN architecture applied to other visual recognition tasks (such as object detection, object localization, and semantic segmentation) is generally derived from the network architecture in image classification. In the wake of these successes, CNN-based methods have emerged in remote sensing image scene classification and achieved advanced classification accuracy. In this review, which focuses on the application of CNNs to image classification tasks, we cover their development, from their predecessors up to recent state-of-the-art (SOAT) network architectures. Along the way, we analyze (1) the basic structure of artificial neural networks (ANNs) and the basic network layers of CNNs, (2) the classic predecessor network models, (3) the recent SOAT network algorithms, (4) comprehensive comparison of various image classification methods mentioned in this article. Finally, we have also summarized the main analysis and discussion in this article, as well as introduce some of the current trends.

A reasonable grasping strategy is a prerequisite for the successful grasping of a target, and it is also a basic condition for the wide application of robots. Presently, mainstream grippers on the market are divided into two-finger grippers and three-finger grippers. According to human grasping experience, the stability of three-finger grippers is much better than that of two-finger grippers. Therefore, this paper’s focus is on the three-finger grasping strategy generation method based on the DeepLab V3+ algorithm. DeepLab V3+ uses the atrous convolution kernel and the atrous spatial pyramid pooling (ASPP) architecture based on atrous convolution. The atrous convolution kernel can adjust the field-of-view of the filter layer by changing the convolution rate. In addition, ASPP can effectively capture multi-scale information, based on the parallel connection of multiple convolution rates of atrous convolutional layers, so that the model performs better on multi-scale objects. The article innovatively uses the DeepLab V3+ algorithm to generate the grasp strategy of a target and optimizes the atrous convolution parameter values of ASPP. This study used the Cornell Grasp dataset to train and verify the model. At the same time, a smaller and more complex dataset of 60 was produced according to the actual situation. Upon testing, good experimental results were obtained.