def draw_predictions_dual(
input: dict,
output: dict,
image_id_key="image_id",
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225),
class_colors=[
(0, 0, 0), # 0=background
(0, 255, 0), # no damage (or just 'building' for localization) (green)
(255, 255, 0), # minor damage (yellow)
(255, 128, 0), # major damage (red)
(255, 0, 0), # destroyed (red)
],
):
images = []
num_images = len(input[image_id_key])
for i, image_id in enumerate(range(num_images)):
image_pre = rgb_image_from_tensor(input[INPUT_IMAGE_PRE_KEY][i], mean,
std)
image_pre = cv2.cvtColor(image_pre, cv2.COLOR_RGB2BGR)
image_post = rgb_image_from_tensor(input[INPUT_IMAGE_POST_KEY][i],
mean, std)
image_post = cv2.cvtColor(image_post, cv2.COLOR_RGB2BGR)
image_pre_gt = image_pre.copy()
image_post_gt = image_post.copy()
localization_target = to_numpy(input[INPUT_MASK_PRE_KEY][i].squeeze(0))
damage_target = to_numpy(input[INPUT_MASK_POST_KEY][i])
image_pre_gt = overlay_image_and_mask(image_pre_gt,
localization_target,
class_colors)
image_post_gt = overlay_image_and_mask(image_post_gt, damage_target,
class_colors)
localization_predictions = to_numpy(
output[OUTPUT_MASK_PRE_KEY][i].squeeze(0).sigmoid() > 0.5).astype(
np.uint8)
damage_predictions = to_numpy(
output[OUTPUT_MASK_POST_KEY][i]).argmax(axis=0)
image_pre = overlay_image_and_mask(image_pre, localization_predictions,
class_colors)
image_post = overlay_image_and_mask(image_post, damage_predictions,
class_colors)
overlay_gt = np.column_stack([image_pre_gt, image_post_gt])
overlay = np.column_stack([image_pre, image_post])
overlay = np.row_stack([overlay_gt, overlay])
overlay = longest_max_size(overlay, 1024, cv2.INTER_LINEAR)
cv2.putText(overlay, str(image_id), (10, 15), cv2.FONT_HERSHEY_PLAIN,
1, (250, 250, 250))
images.append(overlay)
return images
def test_longest_max_size(target):
img = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12]], dtype=np.uint8)
expected = np.array([[2, 3], [6, 7], [10, 11]], dtype=np.uint8)
img, expected = convert_2d_to_target_format([img, expected], target=target)
scaled = F.longest_max_size(img, max_size=3, interpolation=cv2.INTER_LINEAR)
assert np.array_equal(scaled, expected)
def draw_predictions(
input: dict,
output: dict,
image_id_key="image_id",
mean=(0.485, 0.456, 0.406),
std=(0.229, 0.224, 0.225),
class_colors=[
(0, 0, 0), # 0=background
(0, 255,
0), # no damage (or just 'building' for localization) (green)
(255, 255, 0), # minor damage (yellow)
(255, 128, 0), # major damage (red)
(255, 0, 0), # destroyed (red)
(127, 127, 127)
],
max_images=32):
images = []
num_images = len(input[image_id_key])
for i in range(num_images):
image_id = input[INPUT_IMAGE_ID_KEY][i]
image_pre = rgb_image_from_tensor(input[INPUT_IMAGE_KEY][i, 0:3, ...],
mean, std)
image_pre = cv2.cvtColor(image_pre, cv2.COLOR_RGB2BGR)
image_post = rgb_image_from_tensor(input[INPUT_IMAGE_KEY][i, 3:6, ...],
mean, std)
image_post = cv2.cvtColor(image_post, cv2.COLOR_RGB2BGR)
image_pre_gt = image_pre.copy()
image_post_gt = image_post.copy()
damage_target = to_numpy(input[INPUT_MASK_KEY][i])
image_pre_gt = overlay_image_and_mask(image_pre_gt, damage_target,
class_colors)
image_post_gt = overlay_image_and_mask(image_post_gt, damage_target,
class_colors)
damage_predictions = to_numpy(output[INPUT_MASK_KEY][i]).argmax(axis=0)
image_pre = overlay_image_and_mask(image_pre, damage_predictions,
class_colors)
image_post = overlay_image_and_mask(image_post, damage_predictions,
class_colors)
overlay_gt = np.column_stack([image_pre_gt, image_post_gt])
overlay = np.column_stack([image_pre, image_post])
overlay = np.row_stack([overlay_gt, overlay])
overlay = longest_max_size(overlay, 1024, cv2.INTER_LINEAR)
cv2.putText(overlay, str(image_id), (10, 15), cv2.FONT_HERSHEY_PLAIN,
1, (250, 250, 250))
images.append(overlay)
if len(images) >= max_images:
break
return images
def preprocess(image_fname, output_dir, image_size=768):
image = cv2.imread(image_fname)
image = crop_black(image, tolerance=5)
image = longest_max_size(image, max_size=image_size, interpolation=cv2.INTER_CUBIC)
image_id = fs.id_from_fname(image_fname)
dst_fname = os.path.join(output_dir, image_id + '.png')
cv2.imwrite(dst_fname, image)
return
def create_text_image(self, index, etc_text_params=None):
if random.random() <= self.same_text_in_batch_prob:
cur_unicode_list = self.cur_unicode_list
else:
cur_unicode_list = self._create_random_text()
self.cur_text_ing_before = "".join(
[chr(c) for c in cur_unicode_list])
if random.random() <= self.same_font_size_in_batch_prob:
font_size = self.cur_font_size
else:
font_size = random.choice(self.font_size_range)
text = "".join([chr(c) for c in cur_unicode_list])
self.cur_text_ing = text
if random.random() <= self.same_text_params_in_batch_prob:
char_params = self.cur_text_params
else:
char_params = self._get_text_params()
char_params['font_size'] = font_size
self.cur_params = char_params
self.cur_font = self.font_list[index]
# import uuid
# char_params['output_path'] = "../testimage/{}_{}_{}.jpg".format(self.cur_unicode_list[0], index,
# str(uuid.uuid4()))
# text_image_maker.create_text_image(text, self.font_list[index], **char_params)
char_params['output_path'] = None
char_params['auto_chance_color_when_same'] = True
char_params['raise_exception'] = False
char_params['return_mask'] = self.return_mask
# import json
# print(json.dumps(char_params))
# char_params = json.loads('{"color_mode": "RGB", "paddings": {"right": 0.1050272078196586, "top": 0.26090272932922254, "bottom": 0.17015320517900254, "left": 0.25837411687366774}, "text_border": null, "fg_color": [209, 235, 98, 166], "text_italic": false, "font_size": 15, "bg_img_path": "/home/irelin/resource/font_recognition/bgs/294.winterwax-500x500.jpg", "text_shadow": null, "use_img_persp_trans": true, "pos_ratio": [0.3, 0.2], "text_persp_trans_params": [-0.009976076882772873, 0.036253351174196216], "text_rotate": 14, "bg_img_width_ratio": 1.2, "text_blur": 0, "bg_img_height_ratio": 1.1, "use_bg_color": false, "bg_img_scale": 0.5601430892743575, "use_text_persp_trans": true, "use_binarize": false, "img_persp_trans_params": [-0.02296148027430462, 0.004213654695530833], "text_width_ratio": 1.0, "text_gradient": null, "output_path": null, "auto_chance_color_when_same": true, "text_height_ratio": 1.0}')
# if self.use_debug:
# print(os.path.basename(self.font_list[index]), text)
if etc_text_params:
char_params.update(etc_text_params)
img = text_image_maker.create_text_image(text, self.font_list[index],
**char_params)
if self.return_mask:
img, mask = img
if self.use_same_random_crop_in_batch:
height, width = img.shape[:2]
if width >= height:
img = F.smallest_max_size(img,
max_size=self.input_size,
interpolation=cv2.INTER_LINEAR)
else:
img = F.longest_max_size(img,
max_size=self.input_size,
interpolation=cv2.INTER_LINEAR)
pad_width = self.input_size - img.shape[:2][1]
left = pad_width // 2
right = pad_width - left
img = F.pad_with_params(img,
0,
0,
left,
right,
border_mode=cv2.BORDER_CONSTANT,
value=0)
height, width = img.shape[:2]
if width > self.input_size:
last_index = width - self.input_size
start_index = int(self.crop_start_ratio * last_index)
img = img[:, start_index:start_index + self.input_size, :]
if self.transform is not None:
img = self.transform(image=img)['image']
results = [img]
if self.return_mask:
results.append(mask)
if self.return_text:
results.append(text)
if len(results) > 1:
return results
else:
return img
def apply(self, img, interpolation=cv2.INTER_LINEAR, **params):
if max(img.shape[:2]) < self.max_size:
return img
return F.longest_max_size(img, max_size=self.max_size, interpolation=interpolation)
def longest_max_size(img, interpolation=cv2.INTER_LINEAR, **params):
img = F.longest_max_size(img,
max_size=input_size[1],
interpolation=interpolation)
return img