The increasing application scenarios of projectors across various fields have made them indispensable tools. However, achieving optimal display effects in diverse scenarios presents challenges for projectors. In this study, we propose a novel high-resolution projector compensation method, named DHNet, which integrates a channel and spatial attention module into a depthwise separable convolution-based network. Our method effectively captures spatial information and exploits inter-channel associations, thereby enhancing the network’s capability to handle projection distortions. Experimental results demonstrate that our method achieves comparable compensation performance to state-of-the-art methods while effectively reducing the network’s parameters, making it more lightweight.