This work explores approaches that integrate channel attention with spatial attention for zero-shot learning (ZSL). We propose a new ZSL framework that fuses these two attention mechanisms and aims to improve the model’s classification performance by enhancing the discriminative information in feature representations. In particular, the role of channel attention is to strengthen important channel information, while spatial attention enables the model to capture discriminative spatial patterns. The model can better capture and utilize visual and semantic information by integrating these attention mechanisms into the ZSL task. Results from experiments indicate that the integration of attention modules leads to substantial performance gains in ZSL, thereby supporting the effectiveness of our method.