def __call__(self, results):
img, gt_bboxes, gt_labels = [
results[k] for k in ('img', 'gt_bboxes', 'gt_labels')
]
gt_dict = {}
for bbox, label in zip(gt_bboxes, gt_labels):
gt_dict.setdefault(label, []).append(bbox)
if np.random.rand() < self.matting_ratio:
ind = np.random.randint(len(self.normal_groups()))
group_name = self.normal_groups()[ind]
img_ = []
for i in range(5):
img_name = 'imgs_' + group_name + '_{}.jpg'.format(str(i))
gray = cv2.imread(os.path.join(self.normal_path, img_name), 0)
gray = mmcv.imresize_like(gray, img)
raw = img[:, :, i]
bboxes = gt_dict.get(i + 1)
if bboxes is None:
img_.append(gray[..., np.newaxis])
else:
for bbox in bboxes:
xmin = int(bbox[0])
ymin = int(bbox[1])
xmax = int(bbox[2])
ymax = int(bbox[3])
gray[ymin:ymax, xmin:xmax] = cv2.addWeighted(
raw[ymin:ymax, xmin:xmax], self.blend_ratio,
gray[ymin:ymax,
xmin:xmax], 1 - self.blend_ratio, 0)
img_.append(gray[..., np.newaxis])
results['img'] = np.concatenate(img_, axis=-1)
return results
def __call__(self, results):
matting_flag = True if np.random.rand() < self.matting_ratio else False
if matting_flag:
bg_img = mmcv.imread(random.choice(self.bg_img_paths))
bg_img = mmcv.imresize_like(bg_img, results['img'])
for key in results.get('bbox_fields', []):
for box in results[key]:
bg_img[int(box[1]):int(box[3]),
int(box[0]):int(box[2]), :] = results['img'][
int(box[1]):int(box[3]),
int(box[0]):int(box[2]), :]
results['img'] = bg_img
return results
def __call__(self, results):
matting = True if np.random.rand() < self.matting_ratio else False
if matting:
template_no = np.random.randint(len(self.template_list))
template_im_name = self.template_list[template_no]
img_temp = mmcv.imread(template_im_name)
img_temp = mmcv.imresize_like(img_temp, results['img'])
results['concat_img'] = img_temp
for bbox in results['gt_bboxes']:
xmin = int(bbox[0])
ymin = int(bbox[1])
xmax = int(bbox[2])
ymax = int(bbox[3])
beta = np.random.uniform(0.5, 0.8)
img_temp[ymin: ymax, xmin: xmax, :] = cv2.addWeighted(results['img'][ymin: ymax, xmin: xmax, :], beta,
img_temp[ymin: ymax, xmin: xmax, :], 1-beta, 1)
results['img'] = img_temp
else:
results['concat_img'] = None
return results
def __call__(self, results):
if 'matting' not in results:
matting = True if np.random.rand() < self.matting_ratio else False
results['matting'] = matting
if results['matting']:
ind = np.random.randint(len(self.normal_imgs))
normal_name = self.normal_imgs[ind]
normal_img = mmcv.imread(normal_name)
normal_img = mmcv.imresize_like(normal_img, results['img'])
gt_bboxes_ = []
for bbox in results['gt_bboxes']:
xmin = int(bbox[0])
ymin = int(bbox[1])
xmax = int(bbox[2])
ymax = int(bbox[3])
if self.mode == 'fixed':
normal_img[ymin:ymax, xmin:xmax, :] = cv2.addWeighted(
results['img'][ymin:ymax, xmin:xmax, :],
self.blend_ratio, normal_img[ymin:ymax, xmin:xmax, :],
1 - self.blend_ratio, 0)
gt_bboxes_.append(bbox)
elif self.mode == 'random':
h, w = ymax - ymin, xmax - xmin
img_h, img_w, _ = results['img_shape']
xmin = np.random.randint(0, img_w - w)
ymin = np.random.randint(0, img_h - h)
xmax = xmin + w
ymax = ymin + h
normal_img[ymin:ymax, xmin:xmax, :] = cv2.addWeighted(
results['img'][ymin:ymax, xmin:xmax, :],
self.blend_ratio, normal_img[ymin:ymax, xmin:xmax, :],
1 - self.blend_ratio, 0)
gt_bboxes_.append([xmin, ymin, xmax, ymax])
else:
raise NotImplementedError
results['img'] = normal_img
results['gt_bboxes'] = np.array(gt_bboxes_, dtype=np.float32)
return results
'''
MMCV进行图片变换
参考：https://mmcv.readthedocs.io/en/latest/image.html#resize
'''
import mmcv
import numpy as np
img = mmcv.imread("cluo.jpg")
dst_img = mmcv.imread("boy.jpg")
# ============Resize===============
# resize to a given size
mmcv.imresize(img, (1000, 600), return_scale=True)
# resize to the same size of another image
mmcv.imresize_like(img, dst_img, return_scale=False)
# resize by a ratio
mmcv.imrescale(img, 0.5)
# resize so that the max edge no longer than 1000, short edge no longer than 800
# without changing the aspect ratio
mmcv.imrescale(img, (1000, 800))
# =============Rotate==============
# rotate the image clockwise by 30 degrees.
img_ = mmcv.imrotate(img, 30)
# rotate the image counterclockwise by 90 degrees.
img_ = mmcv.imrotate(img, -90)
def test_imresize_like(self):
a = np.zeros((100, 200, 3))
resized_img = mmcv.imresize_like(self.img, a)
assert resized_img.shape == (100, 200, 3)
def __call__(self, results):
img, gt_bboxes, gt_labels = [
results[k] for k in ('img', 'gt_bboxes', 'gt_labels')
]
img_h, img_w, _ = results['img_shape']
if img_h > 1000:
gt_lst = self.all_gt_bboxes()
if np.random.rand() < self.keep_ratio:
gt_bboxes_ = gt_bboxes.tolist()
gt_labels_ = gt_labels.tolist()
else:
ind = np.random.randint(len(self.normal_imgs))
normal_path = self.normal_imgs[ind]
normal_img = mmcv.imread(normal_path)
normal_img = mmcv.imresize_like(normal_img, img)
img = normal_img
gt_bboxes_, gt_labels_ = [], []
defects = random.choices(gt_lst, k=self.max_per_img)
defects = sorted(defects,
key=(lambda x: x[-1] * x[-2]),
reverse=True)
for det in defects:
defect_img = cv2.imread(os.path.join(self.img_path, det[0]))
label = det[1]
x, y, w, h = list(map(int, det[2:]))
defect = defect_img[y:y + h, x:x + w, :]
mode = np.random.randint(3)
# scale
if mode == 0:
defect = self.bbox_scale(defect)
h, w = defect.shape[:2]
# rotate
elif mode == 1:
ctx = x + w * 0.5
cty = y + h * 0.5
corners = np.hstack(
(x, y, x + w, y, x, y + h, x + w, y + h))
h, w, angle = self.bbox_rotate(corners, h, w)
rotated_img = mmcv.imrotate(defect_img,
-angle,
center=(ctx, cty))
x1 = int(ctx - w * 0.5)
y1 = int(cty - h * 0.5)
x2 = x1 + w
y2 = y1 + h
defect = rotated_img[y1:y2, x1:x2, :]
try:
xmin = np.random.randint(self.left_border,
img_w - self.right_border - w)
ymin = np.random.randint(self.top_border,
img_h - self.bottom_border - h)
xmax = xmin + w
ymax = ymin + h
img[ymin:ymax, xmin:xmax, :] = cv2.addWeighted(
defect, 1, img[ymin:ymax, xmin:xmax, :], 0, 0)
gt_bboxes_.append([xmin, ymin, xmax, ymax])
gt_labels_.append(label)
except:
continue
results['img'] = img
results['gt_bboxes'] = np.array(gt_bboxes_, dtype=np.float32)
results['gt_labels'] = np.array(gt_labels_, dtype=np.int64)
return results
