From 567bf98954820394559db6679f210f4f4081b385 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期四, 31 十月 2024 16:22:18 +0800
Subject: [PATCH] fix(model): 调整Codec子模型中的上采样逻辑并修正准确率计算偏移问题

---
 docs/m2met2/Baseline.md |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/docs/m2met2/Baseline.md b/docs/m2met2/Baseline.md
index adc5307..12b6206 100644
--- a/docs/m2met2/Baseline.md
+++ b/docs/m2met2/Baseline.md
@@ -31,4 +31,8 @@
 ## Baseline results
 The results of the baseline system are shown in Table 3. The speaker profile adopts the oracle speaker embedding during training. However, due to the lack of oracle speaker label during evaluation, the speaker profile provided by an additional spectral clustering is used. Meanwhile, the results of using the oracle speaker profile on Eval and Test Set are also provided to show the impact of speaker profile accuracy. 
 
-![baseline_result](images/baseline_result.png)
\ No newline at end of file
+
+|                |SI-CER(%)     |cpCER(%)  |
+|:---------------|:------------:|----------:|
+|oracle profile  |32.72         |42.92      |
+|cluster  profile|32.73         |49.37      |
\ No newline at end of file

--
Gitblit v1.9.1