def get_big_data(self):
big_root = './data/shtu_dataset/original/'
part_A_train = os.path.join(big_root, 'part_A_final/train_data', 'images')
part_A_test = os.path.join(big_root, 'part_A_final/test_data', 'images')
path_sets = [part_A_train, part_A_test]
big_list = []
for path in path_sets:
for img_path in glob.glob(os.path.join(path, '*.jpg')):
mat = io.loadmat(
img_path.replace('.jpg', '.mat').replace('images', 'ground_truth').replace('IMG_', 'GT_IMG_'))
number = mat["image_info"][0, 0][0, 0][1]
if number[0, 0] >= 400:
big_list.append(img_path)
return big_list
def flows_mat_to_np(self):
self.logger.info('Converting MATLAB flows to Numpy')
flows_mat_path = sorted(glob.glob(os.path.join(self.conf.root_path, self.conf.ds_dir,self.conf.frameDir,'TSP_flows','*flow.mat')))
flows_mat = [io.loadmat(flows_mat_path[i]) for i in range(len(flows_mat_path))]
flows_bvx = [flows_mat[i]['flow'][0][0][0] for i in range(len(flows_mat))]
flows_bvy = [flows_mat[i]['flow'][0][0][1] for i in range(len(flows_mat))]
flows_fvx = [flows_mat[i]['flow'][0][0][2] for i in range(len(flows_mat))]
flows_fvy = [flows_mat[i]['flow'][0][0][3] for i in range(len(flows_mat))]
flows_np = dict()
flows_np['bvx'] = np.asarray(flows_bvx).transpose(1,2,0)
flows_np['bvy'] = np.asarray(flows_bvy).transpose(1,2,0)
flows_np['fvx'] = np.asarray(flows_fvx).transpose(1,2,0)
flows_np['fvy'] = np.asarray(flows_fvy).transpose(1,2,0)
file_out = os.path.join(self.conf.precomp_desc_path,'flows.npz')
self.logger.info('Saving optical flows to: ' + file_out)
np.savez(file_out, **flows_np)
return flows_np
def __init__(self, **kwargs):
"""Load the dataset.
kwargs:
root: the root folder of the CUB_200_2011 dataset.
is_training: if true, load the training data. Otherwise, load the
testing data.
crop: if None, does not crop the bounding box. If a real value,
crop is the ratio of the bounding box that gets cropped.
e.g., if crop = 1.5, the resulting image will be 1.5 * the
bounding box area.
target_size: all images are resized to the size specified. Should
be a tuple of two integers, like [256, 256].
version: either '2011' or '2010'.
Note that we will use the python indexing (labels start from 0).
"""
root = kwargs['root']
is_training = kwargs.get('is_training', True)
crop = kwargs.get('crop', None)
target_size = kwargs['target_size']
version = kwargs.get('version', '2011')
if version == '2011':
images = [
line.split()[1]
for line in open(os.path.join(root, 'images.txt'), 'r')
]
boxes = [
line.split()[1:]
for line in open(os.path.join(root, 'bounding_boxes.txt'), 'r')
]
labels = [
int(line.split()[1]) - 1 for line in open(
os.path.join(root, 'image_class_labels.txt'), 'r')
]
birdnames = [
line.split()[1]
for line in open(os.path.join(root, 'classes.txt'), 'r')
]
name_to_id = dict(zip(birdnames, range(len(birdnames))))
split = [
int(line.split()[1]) for line in open(
os.path.join(root, 'train_test_split.txt'), 'r')
]
elif version == '2010':
# we are using version 2010. We load the data to mimic the 2011
# version data format
images = [
line.strip()
for line in open(os.path.join(root, 'lists', 'files.txt'), 'r')
]
boxes = []
# unfortunately, we need to load the boxes from matlab annotations
for filename in images:
matfile = io.loadmat(
os.path.join(root, 'annotations-mat',
filename[:-3] + 'mat'))
left, top, right, bottom = \
[matfile['bbox'][0][0][i][0][0] for i in range(4)]
boxes.append([left, top, right - left, bottom - top])
# get the training and testing split.
train_images = [
line.strip()
for line in open(os.path.join(root, 'lists', 'train.txt'), 'r')
]
labels = [int(line[:line.find('.')]) - 1 for line in images]
birdnames = [
line.strip() for line in open(
os.path.join(root, 'lists', 'classes.txt'), 'r')
]
name_to_id = dict(zip(birdnames, range(len(birdnames))))
split = [int(line in train_images) for line in images]
else:
raise ValueError, "Unrecognized version: %s" % version
# now, do training testing split
if is_training:
target = 1
else:
target = 0
images = [image for image, val in zip(images, split) if val == target]
boxes = [box for box, val in zip(boxes, split) if val == target]
label = [label for label, val in zip(labels, split) if val == target]
# for the boxes, we store them as a numpy array
boxes = np.array(boxes, dtype=np.float32)
boxes -= 1
# load the data
self._data = None
self._load_data(root, images, boxes, crop, target_size)
self._label = np.asarray(label, dtype=np.int)
ndarraydata.NdarrayDataLayer.__init__(
self, sources=[self._data, self._label], **kwargs)
src_folder = '/tmp/jhmdb/'
f = open(filelist, 'r')
gts = []
heights = []
widths = []
preds = []
jnt_visible_set = []
human_boxes = []
feat_res = 40
for cnt, line in enumerate(f):
rows = line.strip().split()
lblpath = rows[1] #+ '/joint_positions.mat'
lbls_mat = sio.loadmat(lblpath)
lbls_coord = lbls_mat['pos_img']
lbls_coord = lbls_coord - 1
gts.append(lbls_coord)
imgpath = rows[0] + '/00001.png'
img = cv2.imread(imgpath)
heights.append(img.shape[0])
widths.append(img.shape[1])
f.close()
# gts = gts[0: 200]
print('read gt')
return labels
#datasets = ['Dataset1','Dataset2','Dataset3','Dataset4','Dataset9','Dataset11','Dataset12','Dataset13','Dataset14','Dataset15','Dataset16','Dataset17','Dataset18','Dataset20']
#datasets = ['Dataset1','Dataset2','Dataset3','Dataset4','Dataset9','Dataset11','Dataset12','Dataset13','Dataset14','Dataset15','Dataset16','Dataset17','Dataset18']
#datasets = ['Dataset15']
datasets = ['Dataset13']
dir_root = '/home/laurent.lejeune/medical-labeling'
dir_frames = 'input-frames'
for i in range(len(datasets)):
old_labels = io.loadmat(
os.path.join(dir_root, datasets[i], dir_frames,
'sp_labels.mat'))['sp_labels']
new_labels = relabel(old_labels)
data = dict()
data['sp_labels'] = new_labels
mat_file_out = os.path.join(dir_root, datasets[i], dir_frames,
'sp_labels.npz')
np.savez(mat_file_out, **data)
sp_desc_df = pd.read_pickle(
os.path.join(dir_root, datasets[i], 'precomp_descriptors',
'sp_desc_df.p'))
##Move old
def __init__(self, **kwargs):
"""Load the dataset.
kwargs:
root: the root folder of the CUB_200_2011 dataset.
is_training: if true, load the training data. Otherwise, load the
testing data.
crop: if None, does not crop the bounding box. If a real value,
crop is the ratio of the bounding box that gets cropped.
e.g., if crop = 1.5, the resulting image will be 1.5 * the
bounding box area.
target_size: all images are resized to the size specified. Should
be a tuple of two integers, like [256, 256].
version: either '2011' or '2010'.
Note that we will use the python indexing (labels start from 0).
"""
root = kwargs['root']
is_training = kwargs.get('is_training', True)
crop = kwargs.get('crop', None)
target_size = kwargs['target_size']
version = kwargs.get('version', '2011')
if version == '2011':
images = [line.split()[1] for line in
open(os.path.join(root, 'images.txt'), 'r')]
boxes = [line.split()[1:] for line in
open(os.path.join(root, 'bounding_boxes.txt'),'r')]
labels = [int(line.split()[1]) - 1 for line in
open(os.path.join(root, 'image_class_labels.txt'), 'r')]
birdnames = [line.split()[1] for line in
open(os.path.join(root, 'classes.txt'), 'r')]
name_to_id = dict(zip(birdnames, range(len(birdnames))))
split = [int(line.split()[1]) for line in
open(os.path.join(root, 'train_test_split.txt'),'r')]
elif version == '2010':
# we are using version 2010. We load the data to mimic the 2011
# version data format
images = [line.strip() for line in
open(os.path.join(root, 'lists', 'files.txt'), 'r')]
boxes = []
# unfortunately, we need to load the boxes from matlab annotations
for filename in images:
matfile = io.loadmat(os.path.join(root, 'annotations-mat',
filename[:-3]+'mat'))
left, top, right, bottom = \
[matfile['bbox'][0][0][i][0][0] for i in range(4)]
boxes.append([left, top, right-left, bottom-top])
# get the training and testing split.
train_images = [line.strip() for line in
open(os.path.join(root, 'lists', 'train.txt'), 'r')]
labels = [int(line[:line.find('.')]) - 1 for line in images]
birdnames = [line.strip() for line in
open(os.path.join(root, 'lists', 'classes.txt'),'r')]
name_to_id = dict(zip(birdnames, range(len(birdnames))))
split = [int(line in train_images) for line in images]
else:
raise ValueError, "Unrecognized version: %s" % version
# now, do training testing split
if is_training:
target = 1
else:
target = 0
images = [image for image, val in zip(images, split) if val == target]
boxes = [box for box, val in zip(boxes, split) if val == target]
label = [label for label, val in zip(labels, split) if val == target]
# for the boxes, we store them as a numpy array
boxes = np.array(boxes, dtype=np.float32)
boxes -= 1
# load the data
self._data = None
self._load_data(root, images, boxes, crop, target_size)
self._label = np.asarray(label, dtype=np.int)
ndarraydata.NdarrayDataLayer.__init__(
self, sources=[self._data, self._label], **kwargs)
currentIndex = 1
#GENERATE STRINGS
for i in range(1,maxIndex+1):
currentSNR = SNRs[SNRIndex]
currentString = 'SEN'+str(i)+'_'+str(currentSNR)+'_'+noiseType+'.mat'
allStrings.append(currentString)
SNRIndex += 1
if (SNRIndex == 9):
SNRIndex = 0
print("LOADING DATA")
AIMImages = list()
for i in range(0,maxIndex):
currentFileLocation = AIMImagesFolder + allStrings[i]
currentTargetLocation = targetsFolder + allStrings[i]
currentFile = io.loadmat(currentFileLocation)
currentFile = np.array(currentFile['AIMImages'])
currentTarget = io.loadmat(currentTargetLocation)
currentTarget = np.transpose(currentTarget['target'])
if(i == 0):
targets = currentTarget
else:
targets = np.concatenate((targets,currentTarget),0)
AIMImages.append(currentFile)
print("FILE IN : " + currentFileLocation + " LOADED")
AIMImages = np.dstack(AIMImages)
print("DATA LOADED")
Xtraining = AIMImages
Xtraining = np.swapaxes(Xtraining,2,1)
Xtraining = np.swapaxes(Xtraining,0,1)