ikergarcia1996 / handwritten-names-recognition Goto Github PK

Hi,

I am following your code and it is working good on your data.
Now I am trying to fit the model on my custom data. I have changed the template and the image and passing the cropped image to the get_characters() function, it is showing a single character.
Please suggests if I am missing something.

Hi,

These are all the list of issues I get when i try to run the source.
Please give some suggestion to solve this error.
Thanks in advance...

Cell : Calculate the Histogram of Oriented Gradients¶
line 360 : fd_train[i], hog_image = hog(X_train_chars[i].reshape(28, 28), orientations=8, pixels_per_cell=(4, 4), cells_per_block=(1, 1), visualise=True)

File "/home/cd005/hulk/lib/python3.6/site-packages/skimage/feature/_hog.py", line 202, in hog
g_row, g_col = _hog_channel_gradient(image)
File "/home/cd005/hulk/lib/python3.6/site-packages/skimage/feature/_hog.py", line 38, in _hog_channel_gradient
g_row[1:-1, :] = channel[2:, :] - channel[:-2, :]
TypeError: numpy boolean subtract, the - operator, is deprecated, use the bitwise_xor, the ^ operator, or the logical_xor function instead.