The anchor-free method based on key point detection has made great progress. However, the anchor-free method is too dependent on using a convolutional network to generate a rough heat map. It is difficult for the network to detect objects whose shape changes greatly, and small objects are difficult to detect. In order to solve this problem, first of all, we use the most advanced mask attention mechanism algorithm in the network to increase the accuracy of the thermodynamic generation of network detection. Then, We also designed an optimized fire model to reduce the size of the model. The masking mechanism optimizes the feature map of the network to enhance the detection capability in network space and improve the accuracy of heatmaps generation. Our approach achieved an accuracy of 91.84% and a recall 89.83% in the Tencent-100K dataset. Our approach is also competitive with the most advanced approach.