pageedward

Results 18 issues of pageedward

I wanna ask as title describe..

how to transform the graph of inference to frozen_graph

what is your loss when ended training?

aster基于attention+seq2seq是不是对长文本效果不好,目前中文训练集普遍长度20+,识别效果很差,一张图几乎没有对的字符,有谁训练中文长文本效果好的嘛,指教下,谢谢

请问aster中文字符训练数据一定需要keypoints,之前用crnn训练的时候只有 ![image](https://user-images.githubusercontent.com/26244036/56780397-b2805b80-6811-11e9-9802-f0e9d1ded781.png)

have anyone train with tf>=1.4, company server cannot allow switch cuda8,

after checkpointsaveHook created , process was killed ![image](https://user-images.githubusercontent.com/26244036/56945560-2ac08700-6b5a-11e9-8da7-ab0a9e6f27ff.png)

File "/home/page/wp/domain-transfer-network/solver.py", line 41, in load_svhn images = np.transpose(svhn['X'], [3, 0, 1, 2]) / 127.5 - 1 MemoryError

使用这份代码在im2latex 100k 上面训练,得到如下的指标:和论文上的Em76相差有些大,anyone meet the same situation? Eval: BLEU-4 88.89 - EM 37.86 - Edit 91.88 - perplexity -1.14

![image](https://user-images.githubusercontent.com/26244036/76440791-58709080-63f9-11ea-8c64-5109d43fb91d.png) ![image](https://user-images.githubusercontent.com/26244036/76440883-7c33d680-63f9-11ea-9dd9-221758bacbbc.png) ![image](https://user-images.githubusercontent.com/26244036/76440899-82c24e00-63f9-11ea-830e-3112f6179cf9.png)