def __new__(cls, root, subset='validation', num_cores=28, transform=None, filter=None):
assert subset in ('validation', 'train'), \
'only support subset (validation, train)'
logger.warning('This api is going to be deprecated, '
'please use ImageRecord instead')
from tensorflow.python.platform import gfile
glob_pattern = os.path.join(root, '%s-*-of-*' % subset)
file_names = gfile.Glob(glob_pattern)
if not file_names:
raise ValueError('Found no files in --root matching: {}'.format(glob_pattern))
from tensorflow.python.data.experimental import parallel_interleave
from lpot.experimental.data.transforms.imagenet_transform import ParseDecodeImagenet
ds = tf.data.TFRecordDataset.list_files(file_names, shuffle=False)
ds = ds.apply(
parallel_interleave(
tf.data.TFRecordDataset, cycle_length=num_cores))
if transform is not None:
transform.transform_list.insert(0, ParseDecodeImagenet())
else:
transform = ParseDecodeImagenet()
ds = ds.map(transform, num_parallel_calls=None)
ds = ds.prefetch(buffer_size=tf.data.experimental.AUTOTUNE) # this number can be tuned
return ds
def result(self):
"""calculate metric"""
if self.num_sample == 0:
logger.warning("sample num is 0 can't calculate topk")
return 0
else:
return self.num_correct / self.num_sample
def iterator_sess_run(sess, iter_op, feed_dict, output_tensor, iteration=-1):
"""Run the graph that have iterator integrated in the graph
Args:
sess (tf.compat.v1.Session): the model sess to run the graph
iter_op (Operator): the MakeIterator op
feed_dict(dict): the feeds to initialize a new iterator
output_tensor(list): the output tensors
iteration(int): iterations to run, when -1 set, run to end of iterator
Returns:
preds: the results of the predictions
"""
sess.run(iter_op, feed_dict)
preds = []
idx = 0
while idx < iteration or iteration == -1:
try:
prediction = sess.run(output_tensor)
preds.append(prediction)
idx += 1
except tf.errors.OutOfRangeError:
break
except:
logger.warning('not run out of the preds...')
break
preds = list(zip(*preds))
return preds
def result(self):
from .coco_tools import COCOWrapper, COCOEvalWrapper
if len(self.ground_truth_list) == 0:
logger.warning("sample num is 0 can't calculate mAP")
return 0
else:
groundtruth_dict = {
'annotations':
self.ground_truth_list,
'images': [{
'id': image_id
} for image_id in self.image_ids],
'categories': [{
'id': k,
'name': v
} for k, v in self.category_map.items()]
}
coco_wrapped_groundtruth = COCOWrapper(groundtruth_dict)
coco_wrapped_detections = coco_wrapped_groundtruth.LoadAnnotations(
self.detection_list)
box_evaluator = COCOEvalWrapper(coco_wrapped_groundtruth,
coco_wrapped_detections,
agnostic_mode=False)
box_metrics, box_per_category_ap = box_evaluator.ComputeMetrics(
include_metrics_per_category=False,
all_metrics_per_category=False)
box_metrics.update(box_per_category_ap)
box_metrics = {
'DetectionBoxes_' + key: value
for key, value in iter(box_metrics.items())
}
return box_metrics['DetectionBoxes_Precision/mAP']
def __init__(self, root, subset='val', num_cores=28, transform=None, filter=None):
self.val_dir = os.path.join(root, subset)
assert os.path.exists(self.val_dir), "find no val dir in {}".format(root) + \
"please make sure there are train/val subfolders"
import glob
logger.warning('This api is going to be deprecated, ' + \
'please use ImageRecord instead')
self.transform = transform
self.image_list = []
files = glob.glob(os.path.join(self.val_dir, '*'))
files.sort()
for idx, file in enumerate(files):
imgs = glob.glob(os.path.join(file, '*'))
for img in imgs:
self.image_list.append((img, idx))
def evaluate(predictions, dataset):
f1 = exact_match = total = 0
for article in dataset:
for paragraph in article['paragraphs']:
for qa in paragraph['qas']:
total += 1
if qa['id'] not in predictions:
message = 'Unanswered question ' + qa['id'] + \
' will receive score 0.'
logger.warning(message)
continue
ground_truths = list(map(lambda x: x['text'], qa['answers']))
prediction = predictions[qa['id']]
f1 += metric_max_over_ground_truths(
f1_score, prediction, ground_truths)
f1 = 100.0 * f1 / total
return f1
def __call__(self, sample):
logger.warning('This transform is going to be deprecated, '
'imagenet decoding will be performed automatically')
return sample