DayBreak-u / chineseocr_lite

超轻量级中文ocr,支持竖排文字识别, 支持ncnn、mnn、tnn推理 ( dbnet(1.8M) + crnn(2.5M) + anglenet(378KB)) 总模型仅4.7M
GNU General Public License v2.0
11.72k stars 2.25k forks source link

检测模型如何训练 #36

Open meta-tabchen opened 4 years ago

meta-tabchen commented 4 years ago

不知道轻量的检测模型是如何训练的?

DayBreak-u commented 4 years ago

我是使用这个项目 https://github.com/WenmuZhou/PSENet.pytorch

xumoremore commented 4 years ago

你好,你的意思是说使用这个链接的代码去训练模型,然后把训练好的模型替换到您的代码中吗。但是这个代码训练完的模型不是有上百MB了吗

xumoremore commented 4 years ago

@ouyanghuiyu

YJLCV commented 3 years ago

@xumoremore 我也想重训练,请问您成功了么

shakyun commented 4 months ago

@xumoremore 你好 训练出来了吗 我也想重新训练 加一些符号