天涯古巷

下载非单文件版的 然后程序放在大容量的硬盘里面，另外新的代码里加入了清除缓存的功能了

繁体中文数据集太小 训练出来的模型效果很差

需要很多打标好的繁体中文数据集来训练，简单来说就是繁体中文的文本图片（样本）和对应的文本内容（标签），然后打标工作量很大，所以支持起来也没这么快

It has the potential to support Vietnamese. I will make it available next iteration.

Could you please provide a single Vietnamese video with me, so that I can test the vietnamese text recognition model. If you could, send it to me: [email protected]

你说的有道理 以后有时间搞一个

This is really a constructive suggestion. This project is so far a small project. The main users are Chinese because the model accuracy in recognizing Chinese language is highest. I...

快速模式抽帧 精准模式逐帧检测 精准的不丢字幕 但是大部分情况下快速也不会丢字幕。原来的逻辑快速用小模型 精准用大模型，后面改为了有大模型就统一优先用大模型

看上去是qpt的问题

I'm so sorry that there is currently no way to support AMD card, for the manufactor do not support CUDA and nvidia is in a monopoly for many deep learning...