diff --git a/ACL_PyTorch/contrib/cv/segmentation/Swin-Transformer-Semantic-Segmentation b/ACL_PyTorch/contrib/cv/segmentation/Swin-Transformer-Semantic-Segmentation
new file mode 100644
index 0000000000000000000000000000000000000000..096f12667aea5fe03082496525e0f04ed49a6259
--- /dev/null
+++ b/ACL_PyTorch/contrib/cv/segmentation/Swin-Transformer-Semantic-Segmentation
@@ -0,0 +1,13 @@
+1.性能对比
+在 310P 设备上，当 batchsize 为 1 时模型性能为 19.33 fps.
+| batchsize | T4性能    | 310P性能    | 310P/T4   |
+|-----------|---------- |----------- |-----------|
+| 1         | 1.72 fps  | 19.33 fps  | 11.23倍     |
+
+注：当 batchsize 为 4 或更高时，因内存不足导致推理失败，无法获取性能数据。
+
+2.精度对比
+自测了 batchsize 为 1 精度，比开源仓精度的高。
+| Model                                  | batchsize | mIoU(NPU) | mIoU(开源)|
+|----------------------------------------|-----------|-----------|----------|
+| Swin-Transformer-Semantic-Segmentation | 1         | 48.06%    | 47.64%   |
\ No newline at end of file