def do_aug(self, data, label, annot):
if self.vis:
self.img_num += 1
# if self.img_num<=self.vis:
# filename = './vis/raw_%d.jpg' % (self.img_num)
# print('save', filename)
# draw = data.copy()
# for i in xrange(label.shape[0]):
# cv2.circle(draw, (label[i][1], label[i][0]), 1, (0, 0, 255), 2)
# scipy.misc.imsave(filename, draw)
rotate = 0
# scale = 1.0
if 'scale' in annot:
scale = annot['scale']
else:
scale = max(data.shape[0], data.shape[1])
if 'center' in annot:
center = annot['center']
else:
center = np.array((data.shape[0] / 2, data.shape[1] / 2))
max_retry = 3
if self.aug_level == 0:
max_retry = 6
retry = 0
found = False
_scale = scale
while retry < max_retry:
retry += 1
succ = True
if self.aug_level > 0:
rotate = np.random.randint(-40, 40)
# rotate2 = np.random.randint(-40, 40)
rotate2 = 0
scale_config = 0.2
# rotate = 0
# scale_config = 0.0
_scale = min(
1 + scale_config,
max(1 - scale_config,
(np.random.randn() * scale_config) + 1))
_scale *= scale
_scale = int(_scale)
# translate = np.random.randint(-5, 5, size=(2,))
# center += translate
if self.mode == 1:
cropped = img_helper.crop2(
data,
center,
_scale, (self.input_img_size, self.input_img_size),
rot=rotate)
if self.use_coherent == 2:
cropped2 = img_helper.crop2(
data,
center,
_scale, (self.input_img_size, self.input_img_size),
rot=rotate2)
else:
cropped = img_helper.crop(
data,
center,
_scale, (self.input_img_size, self.input_img_size),
rot=rotate)
# print('cropped', cropped.shape)
label_out = np.zeros(self.label_shape, dtype=np.float32)
label2_out = np.zeros(self.label_shape, dtype=np.float32)
G = 0
# if self.use_coherent:
# G = 1
_g = G
if G == 0:
_g = 1
# print('shape', label.shape, label_out.shape)
for i in xrange(label.shape[0]):
pt = label[i].copy()
pt = pt[::-1]
# print('before gaussian', label_out[i].shape, pt.shape)
_pt = pt.copy()
trans = img_helper.transform(
_pt,
center,
_scale, (self.output_label_size, self.output_label_size),
rot=rotate)
# print(trans.shape)
if not img_helper.gaussian(label_out[i], trans, _g):
succ = False
break
if self.use_coherent == 2:
_pt = pt.copy()
trans2 = img_helper.transform(
_pt,
center,
_scale,
(self.output_label_size, self.output_label_size),
rot=rotate2)
if not img_helper.gaussian(label2_out[i], trans2, _g):
succ = False
break
if not succ:
if self.aug_level == 0:
_scale += 20
continue
if self.use_coherent == 1:
cropped2 = np.copy(cropped)
for k in xrange(cropped2.shape[2]):
cropped2[:, :, k] = np.fliplr(cropped2[:, :, k])
label2_out = np.copy(label_out)
for k in xrange(label2_out.shape[0]):
label2_out[k, :, :] = np.fliplr(label2_out[k, :, :])
new_label2_out = np.copy(label2_out)
for item in self.mirror_set:
mir = (item[0] - 1, item[1] - 1)
new_label2_out[mir[1]] = label2_out[mir[0]]
new_label2_out[mir[0]] = label2_out[mir[1]]
label2_out = new_label2_out
elif self.use_coherent == 2:
pass
elif ((self.aug_level > 0 and np.random.rand() < 0.5)
or self.force_mirror): # flip aug
for k in xrange(cropped.shape[2]):
cropped[:, :, k] = np.fliplr(cropped[:, :, k])
for k in xrange(label_out.shape[0]):
label_out[k, :, :] = np.fliplr(label_out[k, :, :])
new_label_out = np.copy(label_out)
for item in self.mirror_set:
mir = (item[0] - 1, item[1] - 1)
new_label_out[mir[1]] = label_out[mir[0]]
new_label_out[mir[0]] = label_out[mir[1]]
label_out = new_label_out
if G == 0:
for k in xrange(label.shape[0]):
ind = np.unravel_index(np.argmax(label_out[k], axis=None),
label_out[k].shape)
label_out[k, :, :] = 0.0
label_out[k, ind[0], ind[1]] = 1.0
if self.use_coherent:
ind = np.unravel_index(
np.argmax(label2_out[k], axis=None),
label2_out[k].shape)
label2_out[k, :, :] = 0.0
label2_out[k, ind[0], ind[1]] = 1.0
found = True
break
# self.stats[0]+=1
if not found:
# self.stats[1]+=1
# print('find aug error', retry)
# print(self.stats)
return None
if self.vis > 0 and self.img_num <= self.vis:
print('crop', data.shape, center, _scale, rotate, cropped.shape)
filename = './vis/cropped_%d.jpg' % (self.img_num)
print('save', filename)
draw = cropped.copy()
alabel = label_out.copy()
for i in xrange(label.shape[0]):
a = cv2.resize(alabel[i],
(self.input_img_size, self.input_img_size))
ind = np.unravel_index(np.argmax(a, axis=None), a.shape)
cv2.circle(draw, (ind[1], ind[0]), 1, (0, 0, 255), 2)
scipy.misc.imsave(filename, draw)
if not self.use_coherent:
return cropped, label_out
else:
rotate2 = 0
r = rotate - rotate2
# r = rotate2 - rotate
r = math.pi * r / 180
cos_r = math.cos(r)
sin_r = math.sin(r)
# c = cropped2.shape[0]//2
# M = cv2.getRotationMatrix2D( (c, c), rotate2-rotate, 1)
M = np.array([[cos_r, -1 * sin_r, 0.0], [sin_r, cos_r, 0.0]])
# print(M)
# M=np.array([16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0,
# 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 27, 28, 29, 30, 35, 34, 33,
# 32, 31, 45, 44, 43, 42, 47, 46, 39, 38, 37, 36, 41, 40, 54, 53, 52,
# 51, 50, 49, 48, 59, 58, 57, 56, 55, 64, 63, 62, 61, 60, 67, 66, 65])
return cropped, label_out, cropped2, label2_out, M.flatten()
def get_data(self, data, label, annot):
if self.vis:
self.img_num += 1
#if self.img_num<=self.vis:
# filename = './vis/raw_%d.jpg' % (self.img_num)
# print('save', filename)
# draw = data.copy()
# for i in xrange(label.shape[0]):
# cv2.circle(draw, (label[i][1], label[i][0]), 1, (0, 0, 255), 2)
# scipy.misc.imsave(filename, draw)
rotate = 0
#scale = 1.0
if 'scale' in annot:
scale = annot['scale']
else:
scale = max(data.shape[0], data.shape[1])
if 'center' in annot:
center = annot['center']
else:
center = np.array((data.shape[1] / 2, data.shape[0] / 2))
max_retry = 3
if self.aug_level == 0: #validation mode
max_retry = 6
retry = 0
found = False
base_scale = scale
while retry < max_retry:
retry += 1
succ = True
_scale = base_scale
if self.aug_level > 0:
rotate = np.random.randint(-40, 40)
scale_config = 0.2
#rotate = 0
#scale_config = 0.0
scale_ratio = min(
1 + scale_config,
max(1 - scale_config,
(np.random.randn() * scale_config) + 1))
_scale = int(base_scale * scale_ratio)
#translate = np.random.randint(-5, 5, size=(2,))
#center += translate
data_out, trans = img_helper.transform(data, center,
self.input_img_size,
_scale, rotate)
else:
data, label, trans = img_helper.preprocess(
data, label, data.shape[0])
data_out, trans = img_helper.transform2(
data, label, self.input_img_size)
#data_out = img_helper.crop2(data, center, _scale, (self.input_img_size, self.input_img_size), rot=rotate)
label_out = np.zeros(self.label_shape, dtype=np.float32)
#print('out shapes', data_out.shape, label_out.shape)
for i in xrange(label.shape[0]):
pt = label[i].copy()
#pt = pt[::-1]
npt = img_helper.transform_pt(pt, trans)
if npt[0] >= data_out.shape[1] or npt[1] >= data_out.shape[
0] or npt[0] < 0 or npt[1] < 0:
succ = False
#print('err npt', npt)
break
if config.losstype == 'heatmap':
pt_scale = float(
self.output_label_size) / self.input_img_size
npt *= pt_scale
npt = npt.astype(np.int32)
img_helper.gaussian(label_out[i], npt, config.gaussian)
else:
label_out[i] = (npt / self.input_img_size)
#print('before gaussian', label_out[i].shape, pt.shape)
#trans = img_helper.transform(pt, center, _scale, (self.output_label_size, self.output_label_size), rot=rotate)
#print(trans.shape)
#if not img_helper.gaussian(label_out[i], trans, _g):
# succ = False
# break
if not succ:
if self.aug_level == 0:
base_scale += 20
continue
flip_data_out = None
flip_label_out = None
if config.net_coherent:
flip_data_out, flip_label_out = self.get_flip(
data_out, label_out)
elif ((self.aug_level > 0 and np.random.rand() < 0.5)
or self.force_mirror): #flip aug
flip_data_out, flip_label_out = self.get_flip(
data_out, label_out)
data_out, label_out = flip_data_out, flip_label_out
found = True
break
#self.stats[0]+=1
if not found:
#self.stats[1]+=1
#print('find aug error', retry)
#print(self.stats)
#print('!!!ERR')
return None
#print('found with scale', _scale, rotate)
if self.vis > 0 and self.img_num <= self.vis:
print('crop', data.shape, center, _scale, rotate, data_out.shape)
filename = './vis/cropped_%d.jpg' % (self.img_num)
print('save', filename)
draw = data_out.copy()
alabel = label_out.copy()
for i in xrange(label.shape[0]):
a = cv2.resize(alabel[i],
(self.input_img_size, self.input_img_size))
ind = np.unravel_index(np.argmax(a, axis=None), a.shape)
cv2.circle(draw, (ind[1], ind[0]), 1, (0, 0, 255), 2)
scipy.misc.imsave(filename, draw)
filename = './vis/raw_%d.jpg' % (self.img_num)
scipy.misc.imsave(filename, data)
return data_out, label_out, flip_data_out, flip_label_out
def do_aug(self, data, label, annot):
if self.vis:
self.img_num+=1
#if self.img_num<=self.vis:
# filename = './vis/raw_%d.jpg' % (self.img_num)
# print('save', filename)
# draw = data.copy()
# for i in xrange(label.shape[0]):
# cv2.circle(draw, (label[i][1], label[i][0]), 1, (0, 0, 255), 2)
# scipy.misc.imsave(filename, draw)
rotate = 0
#scale = 1.0
if 'scale' in annot:
scale = annot['scale']
else:
scale = max(data.shape[0], data.shape[1])
if 'center' in annot:
center = annot['center']
else:
center = np.array( (data.shape[0]/2, data.shape[1]/2) )
max_retry = 3
if self.aug_level==0:
max_retry = 6
retry = 0
found = False
_scale = scale
while retry<max_retry:
retry+=1
succ = True
if self.aug_level>0:
rotate = np.random.randint(-40, 40)
#rotate2 = np.random.randint(-40, 40)
rotate2 = 0
scale_config = 0.2
#rotate = 0
#scale_config = 0.0
_scale = min(1+scale_config, max(1-scale_config, (np.random.randn() * scale_config) + 1))
_scale *= scale
_scale = int(_scale)
#translate = np.random.randint(-5, 5, size=(2,))
#center += translate
if self.mode==1:
cropped = img_helper.crop2(data, center, _scale, (self.input_img_size, self.input_img_size), rot=rotate)
if self.use_coherent==2:
cropped2 = img_helper.crop2(data, center, _scale, (self.input_img_size, self.input_img_size), rot=rotate2)
else:
cropped = img_helper.crop(data, center, _scale, (self.input_img_size, self.input_img_size), rot=rotate)
#print('cropped', cropped.shape)
label_out = np.zeros(self.label_shape, dtype=np.float32)
label2_out = np.zeros(self.label_shape, dtype=np.float32)
G = 0
#if self.use_coherent:
# G = 1
_g = G
if G==0:
_g = 1
#print('shape', label.shape, label_out.shape)
for i in xrange(label.shape[0]):
pt = label[i].copy()
pt = pt[::-1]
#print('before gaussian', label_out[i].shape, pt.shape)
_pt = pt.copy()
trans = img_helper.transform(_pt, center, _scale, (self.output_label_size, self.output_label_size), rot=rotate)
#print(trans.shape)
if not img_helper.gaussian(label_out[i], trans, _g):
succ = False
break
if self.use_coherent==2:
_pt = pt.copy()
trans2 = img_helper.transform(_pt, center, _scale, (self.output_label_size, self.output_label_size), rot=rotate2)
if not img_helper.gaussian(label2_out[i], trans2, _g):
succ = False
break
if not succ:
if self.aug_level==0:
_scale+=20
continue
if self.use_coherent==1:
cropped2 = np.copy(cropped)
for k in xrange(cropped2.shape[2]):
cropped2[:,:,k] = np.fliplr(cropped2[:,:,k])
label2_out = np.copy(label_out)
for k in xrange(label2_out.shape[0]):
label2_out[k,:,:] = np.fliplr(label2_out[k,:,:])
new_label2_out = np.copy(label2_out)
for item in self.mirror_set:
mir = (item[0]-1, item[1]-1)
new_label2_out[mir[1]] = label2_out[mir[0]]
new_label2_out[mir[0]] = label2_out[mir[1]]
label2_out = new_label2_out
elif self.use_coherent==2:
pass
elif ((self.aug_level>0 and np.random.rand() < 0.5) or self.force_mirror): #flip aug
for k in xrange(cropped.shape[2]):
cropped[:,:,k] = np.fliplr(cropped[:,:,k])
for k in xrange(label_out.shape[0]):
label_out[k,:,:] = np.fliplr(label_out[k,:,:])
new_label_out = np.copy(label_out)
for item in self.mirror_set:
mir = (item[0]-1, item[1]-1)
new_label_out[mir[1]] = label_out[mir[0]]
new_label_out[mir[0]] = label_out[mir[1]]
label_out = new_label_out
if G==0:
for k in xrange(label.shape[0]):
ind = np.unravel_index(np.argmax(label_out[k], axis=None), label_out[k].shape)
label_out[k,:,:] = 0.0
label_out[k,ind[0],ind[1]] = 1.0
if self.use_coherent:
ind = np.unravel_index(np.argmax(label2_out[k], axis=None), label2_out[k].shape)
label2_out[k,:,:] = 0.0
label2_out[k,ind[0],ind[1]] = 1.0
found = True
break
#self.stats[0]+=1
if not found:
#self.stats[1]+=1
#print('find aug error', retry)
#print(self.stats)
return None
if self.vis>0 and self.img_num<=self.vis:
print('crop', data.shape, center, _scale, rotate, cropped.shape)
filename = './vis/cropped_%d.jpg' % (self.img_num)
print('save', filename)
draw = cropped.copy()
alabel = label_out.copy()
for i in xrange(label.shape[0]):
a = cv2.resize(alabel[i], (self.input_img_size, self.input_img_size))
ind = np.unravel_index(np.argmax(a, axis=None), a.shape)
cv2.circle(draw, (ind[1], ind[0]), 1, (0, 0, 255), 2)
scipy.misc.imsave(filename, draw)
if not self.use_coherent:
return cropped, label_out
else:
rotate2 = 0
r = rotate - rotate2
#r = rotate2 - rotate
r = math.pi*r/180
cos_r = math.cos(r)
sin_r = math.sin(r)
#c = cropped2.shape[0]//2
#M = cv2.getRotationMatrix2D( (c, c), rotate2-rotate, 1)
M = np.array( [
[cos_r, -1*sin_r, 0.0],
[sin_r, cos_r, 0.0]
] )
#print(M)
#M=np.array([16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0,
# 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 27, 28, 29, 30, 35, 34, 33,
# 32, 31, 45, 44, 43, 42, 47, 46, 39, 38, 37, 36, 41, 40, 54, 53, 52,
# 51, 50, 49, 48, 59, 58, 57, 56, 55, 64, 63, 62, 61, 60, 67, 66, 65])
return cropped, label_out, cropped2, label2_out, M.flatten()