This is something small for printing out classification details for a keras ImageDataGenerator .
Image you have three different datagenerators..
train_batches = ImageDataGenerator().flow_from_directory(train_path, target_size=(512,512), classes=klasses, batch_size=32) valid_batches = ImageDataGenerator().flow_from_directory(valid_path, target_size=(512,512), classes=klasses, batch_size=32) test_batches = ImageDataGenerator().flow_from_directory(test_path, target_size=(512,512), classes=klasses, batch_size=32) dataset_stats([train_batches, valid_batches, test_batches], klasses)
Than you can use something like..
def dataset_stats(data_generators=[], klasses=[]):
'''
'''
# collecting data
data = {}
for batch in data_generators:
stats = {}
batch_name = str(batch.directory.split('/')[-1])
for label in batch.labels:
if klasses[label] in stats:
stats[klasses[label]] = stats[klasses[label]]+1
else:
stats[klasses[label]] = 1
data[batch_name] = stats
# printing sth clean
line = '{:20}'.format('Image Counts')
for batch_name in data.keys():
line += '{:>20}'.format(batch_name)
lines = line + '\n'
for klass in klasses:
line = '{:20}'.format(klass)
for batch_name in data.keys():
value = 0
if klass in data[batch_name]:
value = data[batch_name][klass]
line += '{:>20}'.format(value)
lines += '\n' + line
lines += '\n\n{:20}'.format('SUM')
for batch_name in data.keys():
summ = sum(item for item in data[batch_name].values())
lines += '{:>20}'.format(summ)
return lines
To generate a clean output like this..
Image Counts test val train male 1000 1000 1000 female 1000 1000 1000 child 1000 1000 1000 SUM 3000 3000 3000
