In response to the high cost and low efficiency of traditional monitoring methods for the popu-lation of spotted seals in the Bohai and Yellow Seas, this paper proposes a facial feature-based individual recognition model for spotted seals. This model is based on a convolutional neural network architecture, which integrates attention mechanism and multi-scale feature processing technology to enhance the recognition ability of key facial features. It also utilizes depthwise separable convolution to reduce model parameters and computational requirements. The results showed that integrating SlimC2F, CBAM, and SPPF modules significantly enhances network recognition performance, with the proposed method (SlimC2F-CBAM+SPPF) achieving the best results. On the self-built dataset with backgrounds and the Caltech-256 public dataset, the model recognition accuracy reaches 98.35% and 80.70%, outperforming current mainstream methods. With the application of transfer learning strategies, the model's adaptability to different back-grounds has been enhanced, with an increase in identification accuracy to 98.70%. In addition, the model maintains high identification accuracy even on smaller datasets. Finally, visualization analysis demonstrated that the model primarily focused on capturing key areas such as eyes and nose for spotted seal facial recognition, further validating the effectiveness and interpretability of this method while providing new technical means for efficient monitoring of spotted seals.