In order to improve the efficiency of detecting defects on the metal surface of refrigerators and to deal with complex production situations, the Metal-YOLOv3 model was proposed. Using random parameter transformation, the defect data was expanded hundreds of times; the loss function of the original YOLOv3 (You Only Look Once version 3) model was changed, and the Complete Intersection-over-Union (CIoU) loss function based on CIoU was introduced; the threshold of non-maximum suppression algorithm was reduced by using the distribution characteristics of defects; and the anchor value that is more suitable for the data characteristics was calculated based on K-means clustering algorithm, so as to improve the detection accuracy. After a series of experiments, it is found that the Metal-YOLOv3 model is far better than the mainstream Regional Convolutional Neural Network (R-CNN) model in term of detection speed with the Frames Per Second (FPS) reached 7.59, which is 14 times faster than that of Faster R-CNN, and has the Average Precision (AP) reached 88.96%, which is 11.33 percentage points higher than Faster R-CNN, showing the good robustness and generalization performance of the proposed model. It can be seen that this method is effective and can be practically applied to the production of metal products.

For the problem of exponential explosion of joint action space with the increase of the number of agents in multi-agent systems, the "central training-decentralized execution" framework was adopted to solve the curse of dimensionality of joint action space and reduce the optimization cost of the algorithm. A new global credit assignment mechanism, Reward HighWay Network (RHWNet), was proposed to solve the problem that only the global reward corresponding to the joint behavior of all agents was given by the environment in multiple multi-agent reinforcement learning scenarios. By introducing the reward highway connection in the global reward assignment mechanism of the original algorithm, the value function of each agent was directly connected with the global reward, so that each agent was able to consider both the global reward signal and its actual reward value when making strategy selection. Firstly, in the training process, each agent was coordinated through a centralized value function structure. At the same time, this centralized structure was also able to play a role in global reward assignment. Then, the reward highway connection was introduced in the central value function structure to assist the global reward assignment, thus establishing the reward highway network. Then, in the execution phase, each agent's strategy depended only on its own value function. Experimental results on the StarCraft Multi-Agent Challenge (SMAC) microoperation scenarios show that the proposed reward highway network achieves a performance improvement of more than 20% in testing winning rate on four complex maps compared to the advanced Counterfactual multi-agent policy gradient (Coma) and QMIX algorithms. More importantly, in 3s5z and 3s6z scenarios with a large number and different types of agents, the proposed network can achieve better results when the required number of samples is only 30% of algorithms such as Coma and QMIX.

Concerning the problem that face recognition is affected by the factors such as attitude, occlusion, resolution and so on, a method for image super-resolution and face frontalization in unconstrained image was proposed, which could generate high-quality and standard front images. The projection matrix between the input image and 3D model was estimated to generate the standard front image. Also, through the characteristics of face symmetry, the missing pixels by occlusion and attitude could be filled. In order to avoid the loss of pixel information during the process of generating standard front image and improve the image quality, a deeply-recursive convolutional network which had 16 layers was introduced for image super-resolution. To ease the difficulty of training, two extensions were proposed:recursive-supervision and skip-connection. The experimental results on the processed LFW datasets show that it is surprisingly effective when used for face recognition and gender estimation.