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

识别区域太少了,楼主有没有兴趣识别更多区域 #9

Open
wulouzhu opened this issue Oct 31, 2019 · 7 comments
Open

识别区域太少了,楼主有没有兴趣识别更多区域 #9

wulouzhu opened this issue Oct 31, 2019 · 7 comments

Comments

@wulouzhu
Copy link

No description provided.

@helloworld1973
Copy link

cannot agree more

@guanshuicheng
Copy link
Owner

更多区域的检测识别模型会在后期开放

@simplew2011
Copy link

您好,只检测了几个区域,是在yolo训练的时候只用了这几个区域的标注数据吧。

@guanshuicheng
Copy link
Owner

用了这几个区域的标注数据

对,只标注了指定的这五个区域

@simplew2011
Copy link

simplew2011 commented Jun 18, 2020

用了这几个区域的标注数据

对,只标注了指定的这五个区域

【您这五个区域,用yolo训练,最终loss能收敛到多少呢?】

我训练时的一些情况:
基于银行卡数据集https://github.com/cuishuaihua/Bankcard_OCR,
用chineseocr原作者训练代码,YOLO_Kmeans生成anchors = '8,46, 8,57, 8,66, 8,75, 8,84, 8,93, 8,103, 8,115, 8,129',因为只是银行卡数字,长度是一样的,只是原始图像尺寸上存在一些变化,所以聚类生成的anchors变化不大。
将chineseocr的文字检测模型作为resume加载,训练了10轮loss下降很慢
1606/1946 [=======================>......] - ETA: 3:28 - loss: 228.6397 - class_loss_loss: 0.0021 - xy_loss_loss: 79.8738 - wh_loss_loss: 0.2713 - confidence_loss_loss: 131.8300

@Andy546
Copy link

Andy546 commented Dec 14, 2020

用了这几个区域的标注数据

对,只标注了指定的这五个区域

请问下您的数据量大约多少呢?就是标注发票张数和生成发票张数

@yanhaojie569
Copy link

用了这几个区域的标注数据

对,只标注了指定的这五个区域

【您这五个区域,用yolo训练,最终loss能收敛到多少呢?】

我训练时的一些情况: 基于银行卡数据集https://github.com/cuishuaihua/Bankcard_OCR, 用chineseocr原作者训练代码,YOLO_Kmeans生成anchors = '8,46, 8,57, 8,66, 8,75, 8,84, 8,93, 8,103, 8,115, 8,129',因为只是银行卡数字,长度是一样的,只是原始图像尺寸上存在一些变化,所以聚类生成的anchors变化不大。 将chineseocr的文字检测模型作为resume加载,训练了10轮loss下降很慢 1606/1946 [=======================>......] - ETA: 3:28 - loss: 228.6397 - class_loss_loss: 0.0021 - xy_loss_loss: 79.8738 - wh_loss_loss: 0.2713 - confidence_loss_loss: 131.8300

我想问下这个是调模型就好了嘛

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

No branches or pull requests

6 participants