resize function bug

[sevenHsu]

Question describe

• Bug position: data_manage.py->resize_omage function

• code:

    im_arr = imread(image, mode='L')
     r, c = np.shape(im_arr)
     if c > input_width:
         c = input_width
         ratio = float(input_width) / c
         final_arr = imresize(im_arr, (int(32 * ratio), input_width))
     else:
         final_arr = np.zeros((32, input_width))
         ratio = 32.0 / r
         im_arr_resized = imresize(im_arr, (32, int(c * ratio)))
         final_arr[:, 0:np.shape(im_arr_resized)[1]] = im_arr_resized
     return final_arr, c

• detail: if the shape of im_arr is (22,92),the programe will run code in 'else',so the shape of im_arr_resized will be(32,int(c(32/r))) i.e (32,int(92(32/22)))=(32,133).

  At this code : final_arr[:, 0:np.shape(im_arr_resized)[1]] = im_arr_resized.You just make the fina_arr[ : ,0 : 133]=im_arr_resized,but the max_width of final_arr is 100

[Belval]

Just to clarify, you believe that the resulting array should be (32, 100) or (32, 133)?

Using a slice like in the code pictured should produce a (32, 100) array as expected and desired.

[xxlxx1]

@sevenHsu 我看这里的代码也有这个疑惑，test的时候这里就报错了。我改成如果长度太长，只取前面100列，感觉这样还是有问题的，不能识别后面的文字了。 还有一个问题，我把放入session run之前的batchx图片保存下来，它是被reshape了的，宽和高纬度交换的时候，所以保存出来的看不出字的形状，感觉很奇怪啊

[xxlxx1]

@Belval I can't understand your answer, when i try a picture in (22,92), will be (32,133) in your code , then will error in next code. another question, when you prepare training data, from raw_batch_x to batch_x in function "__generate_all_test_batches", you just reshape, i think you should use transpose to change the width and height axis. save the batch_x in this step is not a readable picture anymore.

[Belval]

@xxlxx1 I reproduced the problem. We could simply change the line to:

final_arr[:, 0:min(input_width, np.shape(im_arr_resized)[1]] = im_arr_resized[:, 0:input_width]

Please tell me if it works on your size. I got this result:

Before:

After:

from utils import resize_image
from scipy.misc import imsave

im = resize_image('test3.png', 100)
print(len(im[0]))
imsave('result.png', im[0])

I'm investigating your other question.

[Belval]

For your second question:

There indeed seems to be a big problem with the reshaping, I changed it to np.swapaxes, which gives the desired result.

I'm currently retraining the whole thing, I'll keep you updated.

[sevenHsu]

Thank you!

[Belval]

Please check my last commit, it should fix both the issues mentionned here.

[Belval]

Closing for lack of activity.