diff --git a/TensorFlow/contrib/cv/delf/DELF_ID2024_for_TensorFlow/README.md b/TensorFlow/contrib/cv/delf/DELF_ID2024_for_TensorFlow/README.md
index d504d756e5dae1eac086ea8477afe23e60937d53..4f25588a7ef183dc1684c8487928656a0484b68b 100644
--- a/TensorFlow/contrib/cv/delf/DELF_ID2024_for_TensorFlow/README.md
+++ b/TensorFlow/contrib/cv/delf/DELF_ID2024_for_TensorFlow/README.md
@@ -229,7 +229,7 @@ I0522 20:23:57.981321 281473293996400 eval.py:205] Eval cost in 205.720064878463
 
 ## NPU复现结果
 
-1. 训练性能：每训练500个step，平均耗时为299.35s(GPU:79s)
+1. 训练性能：每训练500个step，平均耗时为82s(GPU:79s)
 2. 训练精度：loss收敛至0.62(GPU:0.58), accuracy收敛至0.72905(GPU:0.7813)
 
 ## 模型固化并转化为om文件