KosukeHao

hindi text render incorrect

8

When I try to use this repo to generate hindi text imgs, it works well. But actually all characters are incorrect because PIL.trutype could only render a simple character-to-glyph but...

about the features.zip

3

Thanks for your work. I wanna know that whether the features in features.zip are extracted from arcface or the resnet-101 trained by yourself? The dataset that I found wasn't in...

一些训练情况的问题

1

感谢作者的付出。最近在尝试检测印刷体文本（区别于场景文本检测） 1. 在10w印刷体文本图片上以batch_size=16训练了7000个iters，loss就从0.6降低到了0.1并且后续降低很慢，是否说明数据多样性不够，已经过拟合了？ 2. 预训练的模型是先在icdar17上训练，然后到ctw1500上微调，微调过程是否有什么需要注意的/技巧？ 问题很多，若能得到作者的解答，十分感谢！