Search

Image matting refers to the estimation of the opacity of foreground objects. It requires correct contours and fine details of foreground objects for the matting results. To better accomplish human image matting tasks, we propose the Cascade Image Matting Network with Deformable Graph Refinement, which can automatically predict precise alpha mattes from single human images without any additional inputs. We adopt a network cascade architecture to perform matting from low-to-high resolution, which corresponds to coarse-to-fine optimization. We also introduce the Deformable Graph Refinement (DGR) module based on graph neural networks (GNNs) to overcome the limitations of convolutional neural networks (CNNs). The DGR module can effectively capture long-range relations and obtain more global and local information to help produce finer alpha mattes. We also reduce the computation complexity of the DGR module by dynamically predicting the neighbors and apply DGR module to higher--resolution features. Experimental results demonstrate the ability of our CasDGR to achieve state-of-the-art performance on synthetic datasets and produce good results on real human images.

In order to improve the control performance of the complex mechatronic system, like the cooling tower in HVAC systems, a data-driven control approach based on model-free deep reinforcement learning method is proposed. The deep Q network (DQN) method, as one of the approaches in reinforcement learning (RL) scheme, can learn and obtain robust feedback control laws by using direct interaction data from the environment. At the same time, radial basis function neural network (RBFNN) is used for the design of deep Q network. Furthermore, the prior knowledge is used to develop an exploration strategy for the RL controller. And the strategy can guide the RL controller to explore the action space, which leads reduction on the training time as compared to traditional DQN controller. At last, the results of simulation show that the DQN controller can achieve lower Integral Absolute Error (IAE) and Integral Square Error (ISE) compared to Proportional integral (PI) controller.