Ruijie Yan

Hi, I get the same problem but the #7 issue is missing now. Can anyone please explain again how to solve the "Segmentation fault (core dumped)" error?

As mentioned by #67 and #42 , the model restoring step is very slow (~10min). I wonder is there any connection between model restoring speed and recognition speed.

> I guess there should be two reasons, > > 1. I am not very sure that the 20ms in our paper is whether tested with beam width 1. >...

你好，没有使用预训练权重，从头训练的

Datasets can be downloaded from their official website. Some GitHub projects also provide their download link, such as [https://github.com/HCIILAB/Scene-Text-Recognition](url)

> hi, if you want to run our code directly, you need to convert LMDB to image and txt files. Or you can modify dataset.py, and read images and labels...

> 你好，请问一下如果训练中文模型的话，它的数据集也需要 'sample' 和 ' .txt ' 这种吗？那它的 'alphabet.txt '这个文件该怎么做呢? 数据集格式以及alphabet.txt和英文都是一样的，alphabet.txt里加入中文字符就行

您好，公布的预训练模型只支持英文。论文中中文实验的训练数据包括合成和真实两部分，合成数据是用[SynthText中文版](https://github.com/JarveeLee/SynthText_Chinese_version)生成的，真实数据来自于[RCTW数据集](https://rctw.vlrlab.net/)。

Hi, this is strange, I have run the model with different random seed and can get similar results. The training data I use is not the LMDB format, I just...

(1) Yes, we simply sampled from the whole data, without considering the ratio from each dataset (2) for MJSynth, please reffer to the [offical site](https://www.robots.ox.ac.uk/~vgg/data/text/#sec-synth), for SynthText, please also reffer...