Fig. 1. Two-stage character pose transfer image generation model

Fig. 2. Improved self-attention mechanism

Fig. 3. Texture refinement generation network

Fig. 4. Residual block with self-attention mechanism

Fig. 5. Perform ablation experiments on the DeepFashion dataset.(a) Sample 1; (b) sample 2; (c) sample 3; (d) sample 4

Fig. 6. Perform ablation experiments on the Market-1501^{dataset. (a) Sample 1; (b) sample 2; (c) sample 3; (d) sample 4}