Skip to content
This repository was archived by the owner on Nov 2, 2024. It is now read-only.

Commit 0a0dc12

Browse files
committedApr 3, 2020
docs(r-cnn): 添加训练结果分析
1 parent a89dbf2 commit 0a0dc12

File tree

3 files changed

+12
-2
lines changed

3 files changed

+12
-2
lines changed
 

‎docs/index.md

+2
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,8 @@
1313
3. 分类器训练
1414
4. 目标检测器实现
1515

16+
**本仓库最终实现一个汽车类别目标检测器**
17+
1618
## 模块构成
1719

1820
1. 区域建议生成:`selectivesearch`算法实现,生成类别独立的区域建议

‎docs/分类器训练.md

+3-1
Original file line numberDiff line numberDiff line change
@@ -128,4 +128,6 @@ val Loss: 1.0424 Acc: 0.9246
128128
remiam negative size: 365403, acc: 0.9981
129129
Training complete in 55m 50s
130130
Best val Acc: 0.924645
131-
```
131+
```
132+
133+
**经过`2`轮训练后,分类器损失就开始收敛**

‎docs/微调模型.md

+7-1
Original file line numberDiff line numberDiff line change
@@ -156,4 +156,10 @@ val Loss: 0.5996 Acc: 0.8718
156156
157157
Training complete in 461m 14s
158158
Best val Acc: 0.873228
159-
```
159+
```
160+
161+
模型在训练过程中呈现过拟合现象,可考虑调整学习率、添加权重衰减以及更换优化器的方式:
162+
163+
1. 学习率从`1e-3`调整为`1e-4`
164+
2. 添加`L2`权重衰减,衰减因子为`1e-4`
165+
3. 使用`Adam`替换`SGD`

0 commit comments

Comments
 (0)
Failed to load comments.