How knowing number of images in flow_from_directory

Question

We can generate image dataset using ImageDataGenerator with flow_from_directory method.

train_datagen = ImageDataGenerator(
    rescale=1./255, #scale images from integers 0-255 to floats 0-1.
    shear_range=0.2,
    zoom_range=0.2, # zoom in or out in images
    horizontal_flip=True) #horizontal flip of images
train_set = train_datagen.flow_from_directory(..)

and that displays:

Found 200 images belonging to 2 classes

I would like to write a loop for to count the number of images on train_set

For image in train_set:
    count = count+1
print(count)

but this does'nt display anything!!

score 23 · Answer 1 · answered Apr 30 '20 at 12:04

23

To get the no. of images, try with the below code.

train_set.samples

answered Apr 30 '20 at 12:04

Biranchi

16,120
23
124
161

1

This is the answer! – Stefan May 05 '20 at 11:23
1

Finally found the answer! – raghus Jul 07 '20 at 23:10
2

Is there a way to get the number of samples for each class? – Riley K Sep 01 '20 at 13:58
1

If case someone need need to know number of labels: len(train_generator.class_indices) – Hardik Raval Apr 14 '21 at 10:17
train_set.n would do too. – Amir Zare Dec 08 '22 at 07:48
1

3 years later, this is still VERY relevant, thank you – user1465073 Jul 07 '23 at 05:34

score 4 · Answer 2 · edited Jan 05 '21 at 21:12

To access the number of samples of your dataset you first must know which type it is:

If you are using the ImageDataGenerator then:

type(train_ds)

will return tensorflow.python.keras.preprocessing.image.DirectoryIterator. In this case you can access the number of samples with:

train_ds.samples

However, if you are creating your creating your dataset using:

train_ds = tf.keras.preprocessing.image_dataset_from_directory( rescale = 1/255. , etc...)

then

type(train_ds)

will return tensorflow.python.data.ops.dataset_ops.BatchDataset which means that you can access the number of samples indirectly with

len(train_ds.file_paths)

score 0 · Answer 3 · answered Apr 30 '20 at 12:04

__iter__ will be your solution.

first you try to predict type of train_set .if you do not know.

print(type(train_set))
#then you find **keras.preprocessing.image.DirectoryIterator**

Now you want to apply some function or modification on this type data follow this linkhere

How knowing number of images in flow_from_directory

3 Answers3