Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

基于MindOCR的车牌识别 #760

Merged
merged 23 commits into from
Nov 22, 2024
Merged

基于MindOCR的车牌识别 #760

merged 23 commits into from
Nov 22, 2024

Conversation

Xv-M-S
Copy link
Contributor

@Xv-M-S Xv-M-S commented Oct 30, 2024

Thank you for your contribution to the MindOCR repo.
Before submitting this PR, please make sure:

Motivation

【开源实习】基于MindOCR的车牌识别案例输出 (https://gitee.com/mindspore/community/issues/I98K8W)
为基于MindOCR的车牌号识别案例提供英文readme

Test Plan

(How should this PR be tested? Do you require special setup to run the test or repro the fixed bug?)

Related Issues and PRs

【开源实习】基于MindOCR的车牌识别案例输出 (https://gitee.com/mindspore/community/issues/I98K8W)
(Is this PR part of a group of changes? Link the other relevant PRs and Issues here. Use https://help.github.com/en/articles/closing-issues-using-keywords for help on GitHub syntax)

Copy link
Collaborator

@panshaowu panshaowu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

建议修订部分内容

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 1, 2024

已按照批注修改

@alien-0119
Copy link
Contributor

examples/license_plate_detection_and_recognition/generate_dict.py
针对这个生成字典的py脚本,想请问下,line 106这里加了index是有什么初衷嘛。实际上这么执行后生成的字典似乎不能用,去掉index之后的是否才是正确的呢?

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 5, 2024

是的,您是正确的。这应该是之前为了统计合并后的字符数加上去的,应该去掉。

Copy link
Collaborator

@kk928290341 kk928290341 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. 文档有新要求,麻烦统一按照该样例整理中英文文档的环境信息,性能表现 https://github.com/mindspore-lab/mindone/wiki/%E6%A8%A1%E5%9E%8B%E7%89%88%E6%9C%AC%E4%B9%A6%E5%86%99%E6%A0%B7%E4%BE%8B

  2. 数据集地址直接放官方链接,避免填写网盘地址

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 7, 2024

  1. 文档有新要求,麻烦统一按照该样例整理中英文文档的环境信息,性能表现 https://github.com/mindspore-lab/mindone/wiki/%E6%A8%A1%E5%9E%8B%E7%89%88%E6%9C%AC%E4%B9%A6%E5%86%99%E6%A0%B7%E4%BE%8B
  2. 数据集地址直接放官方链接,避免填写网盘地址

老师,我这用GPU训的没有ascend+cann的版本信息,我写GPU dirver和GPU型号,以及cudatoolkit版本可以不

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 7, 2024

  1. 文档有新要求,麻烦统一按照该样例整理中英文文档的环境信息,性能表现 https://github.com/mindspore-lab/mindone/wiki/%E6%A8%A1%E5%9E%8B%E7%89%88%E6%9C%AC%E4%B9%A6%E5%86%99%E6%A0%B7%E4%BE%8B
  2. 数据集地址直接放官方链接,避免填写网盘地址

1.已添加环境信息和性能表现,请导师审核
2.数据集论文作者只提供了百度网盘的地址和谷歌网盘的地址

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 11, 2024

已按照检视要求修改

@panshaowu
Copy link
Collaborator

  1. 文档有新要求,麻烦统一按照该样例整理中英文文档的环境信息,性能表现 https://github.com/mindspore-lab/mindone/wiki/%E6%A8%A1%E5%9E%8B%E7%89%88%E6%9C%AC%E4%B9%A6%E5%86%99%E6%A0%B7%E4%BE%8B
  2. 数据集地址直接放官方链接,避免填写网盘地址

1.已添加环境信息和性能表现,请导师审核 2.数据集论文作者只提供了百度网盘的地址和谷歌网盘的地址
此处的建议,是直接提供数据集官方github链接,并告诉用户,按照官方github项目链接的指引进行下载,而不是绕过其链接,直接贴百度网盘和GoogleDrive地址。

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 15, 2024

已按照检视要求修改,请导师审核

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 20, 2024 via email

@panshaowu
Copy link
Collaborator

潘导您好! 已按照所提修改意见对文档进行修改,请审核!

@Xv-M-S 你好,请注意上面还有4个处于“pending”状态的review意见。已经当面请第二位reviewer @chengtianwu 补充检视和合入PR了。
请重点关注,当前文档中配图为安装GPU版本的MindSpore,但是提供的环境配套信息和测试数据都是NPU版本(上次修改删去了GPU相关信息),可能会误导用户。建议参考review意见调整下中英文档。

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 21, 2024 via email

@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 21, 2024 via email

@@ -414,7 +418,7 @@ valid res:
[2024-09-10 23:01:26] mindocr.eval INFO - Performance: {'acc': 0.9795991778373718, 'norm_edit_distance': 0.995379626750946}
```

6. 修改`SVTRRecAug`​的`aug_type`​为1:应用更强的数据集增强方式。[svtr_transform.py](https://github.com/mindspore-lab/mindocr/blob/5fd78b46b42d40aeba01f72538699837594053b1/mindocr/data/transforms/svtr_transform.py#L354)
6. 增加数据增强强度:将​配置文件`SVTRRecAug`部分​的`aug_type`修改为1

```java
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

java -> text

@@ -396,7 +400,7 @@ valid res:
[2024-09-10 19:14:02] mindocr.eval INFO - Performance: {'acc': 0.069402776658535, 'norm_edit_distance': 0.765773355960846}
```

4. Resize策略:不考虑宽高比,直接将所有文本图像`resize`​到`32 * 100`​`Resize`​时不使用`Padding`​`max_text_length: 25`​
4. Resize策略: 直接将所有文本图像`resize`​到`32 * 100`​,`Resize`​时不使用`Padding`​;`max_text_length`设置为25​;

```java
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

java -> text


```java
valid res:
[2024-09-10 15:16:38] mindocr.metrics.rec_metrics INFO - correct num: 23, total num: 99996.0
[2024-09-10 15:16:38] mindocr.eval INFO - Performance: {'acc': 0.00023000920191407204, 'norm_edit_distance': 0.5451045036315918}
```

3. 将`image_shape`​调整为【32,80】等比例放缩
3. 修改配置文件`model`部分的`img_size`为[32,80]

```java
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

java -> text

   ├── images
   │   ├── img_1.jpg
   │   ├── img_2.jpg
   │   └── ....jpg
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

....jpg -> ...

@mindspore-lab mindspore-lab deleted a comment from chengtianwu Nov 22, 2024
@mindspore-lab mindspore-lab deleted a comment from chengtianwu Nov 22, 2024
@panshaowu panshaowu merged commit 1fe6c8d into mindspore-lab:main Nov 22, 2024
2 checks passed
@Xv-M-S
Copy link
Contributor Author

Xv-M-S commented Nov 22, 2024 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants