CODEWITHSUNDEEP
pythonkerasconv-neural-network
user2458922
user2458922

Reputation: 1721

Why the CNN is executing more than a training size

why is when len(trainData) = 75 The execution Verbose showing a number greater than 75 ?

My Model Execution Script

batch_size=64
h5_path = "EPOC_1_Feb_25_model.h5"
checkpoint = ModelCheckpoint(h5_path, monitor='val_acc', verbose=1, save_best_only=True, mode='max')

history = model.fit_generator(
    data_gen(train2, id_label_map, batch_size, augment=True),
    validation_data=data_gen(val, id_label_map, batch_size),
    epochs=1, verbose=1,
    callbacks=[checkpoint],
    steps_per_epoch=len(train) // batch_size,
    validation_steps=len(val) // batch_size)
model.load_weights(h5_path)

The Execution Verbose

Epoch 1/1

7658/9409 [=======================>......]

given len(train2) = 75 Why is this showing 7658/9409 ?

My Data Generator is

def data_gen(list_files, id_label_map, batch_size, augment=False):
    seq = get_seq()
    while True:
        shuffle(list_files)
        for batch in chunker(list_files, batch_size):
            X = [cv2.imread(x) for x in batch ] 
            Y = [id_label_map.get(x) for x in batch]#[id_label_map[get_id_from_file_path(x)] for x in batch]
            if augment:
                X = seq.augment_images(X)
            X = [preprocess_input(x) for x in X]
                
            yield np.array(X), np.array(Y)

def chunker(seq, size):
    return (seq[pos:pos + size] for pos in range(0, len(seq), size))

Upvotes: 0

Views: 39

Answers (2)

Gokulaselvan
Gokulaselvan

Reputation: 138

In steps_per_epoch = len(train) //batch_size instead of steps_per_epoch = len(train2) //batch_size Because of this you are getting more numbers in data I guess

Upvotes: 1

Thomas Schillaci
Thomas Schillaci

Reputation: 2453

I think it comes from your data augmentation. Try setting:

augment=False

In your data_gen call to test it.

Upvotes: 0

Related Questions