pickle_dataset.py about facialemotionrecognition

[mahakal]

Do you have the cohn kanade dataset ???
Want seems to be the problem.

Have you checked
https://github.com/mahakal/FacialEmotionRecognition#usage
https://github.com/mahakal/FacialEmotionRecognition/blob/master/pickle_dataset.py#L123

[sarmadm]

HI

Thanks for replying .

Yes I have the extended CK data which is arranged as several folders and inside each folder there is

types of emotions as some of them has labels and others not.

How should I make the pickle_dataset.py to read through these folders and then use cnn.py ?

[mahakal]

https://github.com/mahakal/FacialEmotionRecognition/blob/master/pickle_dataset.py#L119 => put the absolute path of CK dataset.

open python terminal in the repo

from pickle_dataset import PickleData
p = PickleData()
p.start_pickling()

I might refactor the code so it takes care of the directory path, threshold, error handling, etc. But then again it's just pickling the data.

[sarmadm]

Hi

Images bath is : C:/Users/sarmad/Documents/CK/cohn-kanade-images/

where all folders are . I put this path in pickle_dataset.py and run the code in python terminal.

there is a file saved as .p but when I run cnn.py ,

Using TensorFlow backend. Traceback (most recent call last): File "cnn.py", line 34, in <module> X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) ValueError: cannot reshape array of size 0 into shape (8,100,100,1)

[mahakal]

https://github.com/mahakal/FacialEmotionRecognition/blob/master/cnn.py#L25

Replace the name with the generated file.
I'm gonna rewrite it.

[sarmadm]

I renamed and replaced the file name but I get errors I showed in last message . I think there is some problems with reading images from the folders

[sarmadm]

When I load and print pickle file this is the output :
((array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)), (array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)), (array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)))

[mahakal]

@sarmadm Yeah, I'm rewriting the script making it more user-friendly. I'll be done by tomorrow.

[sarmadm]

Thanks , I tried running pickle.py is shows this error

pickle_dataset.py: error: the following arguments are required: C:/Users/Documents/CK/cohn-kanade-images

This the folder that contains all the folders with images as shown in above image , how should I wrtire the path ?

[sarmadm]

Hi

I run the code again shows this error :

X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) ValueError: cannot reshape array of size 0 into shape (8,100,100,1)

is that the case with you ?

[mahakal]

There were some typos, try it again.

For Help:
python pickle_dataset.py -h
python cnn.py -h

Let me know whether it works or not.
And can you tell me, for which purpose are you using the repository.

[sarmadm]

It shows same error , is it problem with ck data folder structure ?

(keras_v) C:\Users\Documents\CK>python cnn.py C:\Users\Documents\CK\ck_dataset.pickle Using TensorFlow backend. Traceback (most recent call last): File "cnn.py", line 55, in <module> X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) ValueError: cannot reshape array of size 0 into shape (8,100,100,1)

[mahakal]

The pickle file seems to be empty. Delete the previous pickle.py file.
Run, python pickle_dataset.py C:/Users/Documents/CK/cohn-kanade-images --crop
The pickle would be created in repo directory.
Then run, python cnn.py ck_dataset.pickle

Let me know whether it works or not.
And can you tell me, for which purpose are you using the repository.

[sarmadm]

I deteted the previous pickle file , and run :

(keras_v) C:\Users\Documents\CK>python pickle_dataset.py C:\Users\Documents\CK\cohn-kanade-images

C:\Users\Documents\CK\cohn-kanade-images folder contains imaged sequence folders

pickle file is saved and then I run :

(keras_v) C:\Users\Documents\CK>python cnn.py C:\Users\Documents\CK\ck_dataset.pickle Using TensorFlow backend. Traceback (most recent call last): File "cnn.py", line 55, in <module> X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1) ValueError: cannot reshape array of size 0 into shape (8,100,100,1)

[sarmadm]

Any idea ??

[mahakal]

There was a bug in the code, fetch the latest and check.
And also, to check the pickle file you could

import pickle
o = pickle.load(open('name_of_the_file.pickle' 'rb'))
o.keys()
print(o)

[sarmadm]

It shows same error and this is ck_dataset.pickle output :

dict_keys(['training_data', 'validation_data', 'test_data', 'img_dim']) print(o)

{training_data': [array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)], 'validation_data': [array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)], 'test_data': [array([], shape=(8, 0), dtype=float64), array([], shape=(0, 1), dtype=int8)], 'img_dim': {'width': 100, 'height': 100}}

[mahakal]

Are you using CK database or Extended CK database???
Try again, there was a bug now the script works for CK database.

[sarmadm]

I'm using Extended CK database , is the code will not work for it ?