def parse_multi_file(filen, ret=None, groundtruth=None):
file = open(filen, "r")
if ret == None:
ret = DataSetMulti()
if drop_neg:
negs_files = [f[:f.rfind(".")] for f in os.listdir(negs)]
i = 0
for line in file:
line = line.strip().rstrip()
splited = line.split()
filename = splited[0]
# filename = filename[filename.rfind("/")+1:]
# filename = filename[:filename.rfind(".")]
if drop_neg:
if filename in negs_files:
continue
height = int(splited[1])
width = int(splited[2])
class_id = int(splited[3])
(confidence, x, y, x2, y2) = tuple([float(a) for a in splited[4:]])
if hratio != None:
height = y2 - y
height2 = height * hratio
y += (height - height2) / 2.0
y2 = y + height2
if wratio != None:
width = x2 - x
width2 = width * wratio
x += (width - width2) / 2.0
x2 = x + width2
if whratio != None:
height = y2 - y
width = x2 - x
width2 = height * whratio
x += (width - width2) / 2.0
x2 = x + width2
bb = BoundingBox(x, y, x2, y2, confidence)
if groundtruth != None:
img = groundtruth[filename]
if img == []:
print "warning: " + filename + " not found in groundtruth."
continue
# r = bb.area() / (img.height * img.width)
# if (min_area_ratio != None and r < min_area_ratio) or \
# (max_area_ratio != None and r > max_area_ratio):
# print "not adding ratio " + str(r)
# continue
area = bb.area()
if (min_area == None or area >= min_area) and \
(max_area == None or area <= max_area):
ret.add_obj(confidence, filename, bb)
i = i + 1
file.close()
return ret
def parse_multi_file(filen, ret = None, groundtruth = None):
file = open(filen, "r")
if ret == None:
ret = DataSetMulti()
if drop_neg:
negs_files = [f[:f.rfind(".")] for f in os.listdir(negs)]
i = 0
for line in file:
line = line.strip().rstrip()
splited = line.split()
filename = splited[0]
# filename = filename[filename.rfind("/")+1:]
# filename = filename[:filename.rfind(".")]
if drop_neg:
if filename in negs_files:
continue
height = int(splited[1])
width = int(splited[2])
class_id = int(splited[3])
(confidence, x, y, x2, y2) = tuple([float(a) for a in splited[4:]])
if hratio != None:
height = y2 - y
height2 = height * hratio
y += (height - height2) / 2.0
y2 = y + height2
if wratio != None:
width = x2 - x
width2 = width * wratio
x += (width - width2) / 2.0
x2 = x + width2
if whratio != None:
height = y2 - y
width = x2 - x
width2 = height * whratio
x += (width - width2) / 2.0
x2 = x + width2
bb = BoundingBox(x, y, x2, y2, confidence)
if groundtruth != None:
img = groundtruth[filename]
if img == []:
print "warning: " + filename + " not found in groundtruth."
continue ;
# r = bb.area() / (img.height * img.width)
# if (min_area_ratio != None and r < min_area_ratio) or \
# (max_area_ratio != None and r > max_area_ratio):
# print "not adding ratio " + str(r)
# continue
area = bb.area()
if (min_area == None or area >= min_area) and \
(max_area == None or area <= max_area):
ret.add_obj(confidence, filename, bb)
i = i + 1
file.close()
return ret
def get_object(data, i):
xmin, xmax, ymin, ymax = None, None, None, None
object_char = data.find("object %d" % (i, ))
while object_char != -1:
begin_line = data[:object_char].rfind("\n") + 1
end_line = begin_line + 1 + data[begin_line + 1:].find("\n")
line = data[begin_line:end_line]
if line.find("Bounding box") != -1:
# if this happens, there are other objects than pedestrians
# in the annotation :
assert (line.find("PASperson") != -1)
box_begin = line.rfind(":") + 1
box_str = line[box_begin:]
box_str = box_str[box_str.find("(") + 1:].strip()
xmin = int(box_str[:box_str.find(",")].strip().rstrip())
box_str = box_str[box_str.find(",") + 1:].strip()
ymin = int(box_str[:box_str.find(")")].strip().rstrip())
box_str = box_str[box_str.find("(") + 1:].strip()
xmax = int(box_str[:box_str.find(",")].strip().rstrip())
box_str = box_str[box_str.find(",") + 1:].strip()
ymax = int(box_str[:box_str.find(")")].strip().rstrip())
break
object_char = object_char+1 \
+ data[object_char+1:].find("object %d"%(i,))
if xmin == None:
return None
if whratio != None:
height = ymax - ymin
width = xmax - xmin
width2 = height * whratio
xmin += (width - width2) / 2.0
xmax = xmin + width2
return BoundingBox(xmin, ymin, xmax, ymax, 1.0)
def on_button_release(self, widget, event):
if event.button == 1:
self.button1 = False
o = self.button1_origin
w0 = max(0, min(o[1], event.x))
h0 = max(0, min(o[0], event.y))
w1 = max(w0, max(o[1], event.x))
h1 = max(h0, max(o[0], event.y))
self.boxes.append(BoundingBox(w0, h0, w1, h1))
self.displayer.draw(self.boxes)
def parse(filen, crawl=False):
file = open(filen, "r")
ret = DataSet()
for line in file:
line = line.strip().rstrip()
splited = line.split()
filename = splited[0]
# filename = filename[filename.rfind("/")+1:]
# filename = filename[:filename.rfind(".")]
height = int(splited[1])
width = int(splited[2])
class_id = int(splited[3])
(confidence, x, y, x2, y2) = tuple([float(a) for a in splited[4:]])
#if confidence > parse_confidence_min: #TODO
if hratio != None:
height = y2 - y
height2 = height * hratio
y += (height - height2) / 2.0
y2 = y + height2
if wratio != None:
width = x2 - x
width2 = width * wratio
x += (width - width2) / 2.0
x2 = x + width2
if whratio != None:
height = y2 - y
width = x2 - x
width2 = height * whratio
x += (width - width2) / 2.0
x2 = x + width2
bb = BoundingBox(x, y, x2, y2)
area = bb.area()
if (min_area == None or area >= min_area) and \
(max_area == None or area <= max_area):
ret.add_obj(filename, bb)
file.close()
# print summary
print 'Dataset ' + str(filen) + ' has ' + str(len(ret)) + ' images and ' \
+ str(ret.get_nobjs()) + ' positive objects.'
return ret
def parse(filen, crawl = False):
file = open(filen, "r")
ret = DataSet()
for line in file:
line = line.strip().rstrip()
splited = line.split()
filename = splited[0]
# filename = filename[filename.rfind("/")+1:]
# filename = filename[:filename.rfind(".")]
height = int(splited[1])
width = int(splited[2])
class_id = int(splited[3])
(confidence, x, y, x2, y2) = tuple([float(a) for a in splited[4:]])
#if confidence > parse_confidence_min: #TODO
if hratio != None:
height = y2 - y
height2 = height * hratio
y += (height - height2) / 2.0
y2 = y + height2
if wratio != None:
width = x2 - x
width2 = width * wratio
x += (width - width2) / 2.0
x2 = x + width2
if whratio != None:
height = y2 - y
width = x2 - x
width2 = height * whratio
x += (width - width2) / 2.0
x2 = x + width2
bb = BoundingBox(x, y, x2, y2)
area = bb.area()
if (min_area == None or area >= min_area) and \
(max_area == None or area <= max_area):
ret.add_obj(filename, bb)
file.close()
# print summary
print 'Dataset ' + str(filen) + ' has ' + str(len(ret)) + ' images and ' \
+ str(ret.get_nobjs()) + ' positive objects.'
return ret
def paste_foreground_into_background(foreground_img, background_img, x_factor=0.5, y_factor=0.5, bounding_box=True):
"""
Paste a foreground/object image onto a background image, translated by the given x and y factors. If parameter
'bounding_box' is true, then also the bounding box is returned.
:param foreground_img: foreground/object image
:param background_img: background image
:param x_factor: value for x translation
:param y_factor: value for y translation
:param bounding_box: boolean, whether bounding box should be returned
:return: merged_img, (bounding_box): merged image and optional the bounding box of the foreground object
"""
# if foreground image is bigger than background image, then adjust foreground's resolution to background's
if foreground_img.shape[0] > background_img.shape[0]:
foreground_img = crop_image(foreground_img, new_height=background_img.shape[0])
if foreground_img.shape[1] > background_img.shape[1]:
foreground_img = crop_image(foreground_img, new_width=background_img.shape[1])
merged_img = background_img.copy()
fg_img_height, fg_img_width = foreground_img.shape[:2]
bg_img_height, bg_img_width = background_img.shape[:2]
x_offset = int((bg_img_width - fg_img_width) * x_factor)
y_offset = int((bg_img_height - fg_img_height) * y_factor)
# calc corners of foreground image
y1, y2 = y_offset, y_offset + foreground_img.shape[0]
x1, x2 = x_offset, x_offset + foreground_img.shape[1]
# calc alpha value (opacity)
alpha_s = foreground_img[:, :, 3] / 255.0
alpha_l = 1.0 - alpha_s
# do actual merge of fore- in background image
for c in range(0, 3):
merged_img[y1:y2, x1:x2, c] = (alpha_s * foreground_img[:, :, c] +
alpha_l * background_img[y1:y2, x1:x2, c])
logging.info(" => merge object and background")
if bounding_box:
return merged_img, BoundingBox(x_min=x1, x_max=x2, y_min=y1, y_max=y2)
else:
return merged_img
def parse_file_multi(file, filename, dataset):
for line in file:
(x, y, w, h, score) = \
tuple([float(a) for a in line.strip().rstrip().split(",")])
if filename not in corresp:
print 'warning: image ' + filename + ' not found in groundtruth'
else:
if hratio != None:
h2 = h * hratio
y += (h - h2) / 2.0
h = h2
if wratio != None:
w2 = w * wratio
x += (w - w2) / 2.0
w = w2
if whratio != None:
w2 = h * whratio
x += (w - w2) / 2.0
w = w2
dataset.add_obj(score, corresp[filename], \
BoundingBox(x, y, x+w, y+h, score))
