def visualize_image(image, boxes, classes, scores, id_mapping):
"""Visualizes a list of images.
Args:
image: a image with shape [H, W, C].
boxes: a box prediction with shape [N, 4] ordered [ymin, xmin, ymax, xmax].
classes: a class prediction with shape [N].
scores: A list of float value with shape [N].
id_mapping: a dictionary from class id to name.
Returns:
output_image: an output image with annotated boxes and classes.
"""
category_index = {k: {'id': k, 'name': id_mapping[k]} for k in id_mapping}
img = np.array(image)
vis_utils.visualize_boxes_and_labels_on_image_array(
img,
boxes,
classes,
scores,
category_index,
min_score_thresh=0.5,
instance_masks=None,
use_normalized_coordinates=False,
line_thickness=4)
return img
def visualize_image(image,
boxes,
classes,
scores,
id_mapping=None,
min_score_thresh=0.01,
max_boxes_to_draw=1000,
line_thickness=2,
**kwargs):
"""Visualizes a given image.
Args:
image: a image with shape [H, W, C].
boxes: a box prediction with shape [N, 4] ordered [ymin, xmin, ymax, xmax].
classes: a class prediction with shape [N].
scores: A list of float value with shape [N].
id_mapping: a dictionary from class id to name.
min_score_thresh: minimal score for showing. If claass probability is below
this threshold, then the object will not show up.
max_boxes_to_draw: maximum bounding box to draw.
line_thickness: how thick is the bounding box line.
**kwargs: extra parameters.
Returns:
output_image: an output image with annotated boxes and classes.
"""
id_mapping = parse_label_id_mapping(id_mapping)
category_index = {k: {'id': k, 'name': id_mapping[k]} for k in id_mapping}
img = np.array(image)
#changes
indices = np.argwhere((classes ==31) | (classes==33) | (classes==44) |
(classes==47) | (classes==48) | (classes==49) | (classes==50) | (classes==51)
| (classes==74)| (classes==75)| (classes==77)| (classes==84)| (classes==86)
| (classes==87)| (classes==88)| (classes==89)| (classes==90))
boxes = np.squeeze(boxes[indices], axis=1) # to prevent errors made by nd.array of size 1 nd.array
scores = np.squeeze(scores[indices], axis=1)
classes = np.squeeze(classes[indices], axis=1)
#ends
vis_utils.visualize_boxes_and_labels_on_image_array(
img,
boxes,
classes,
scores,
category_index,
min_score_thresh=min_score_thresh,
max_boxes_to_draw=max_boxes_to_draw,
line_thickness=line_thickness,
**kwargs)
return img
def test_visualize_boxes_and_labels_on_image_array(self):
ori_image = np.ones([360, 480, 3], dtype=np.int32) * 255
test_image = np.ones([360, 480, 3], dtype=np.int32) * 255
detections = np.array([[0.8, 0.1, 0.9, 0.1, 1., 0.1],
[0.1, 0.3, 0.8, 0.7, 1., 0.6]])
labelmap = {1: {'id': 1, 'name': 'cat'}, 2: {'id': 2, 'name': 'dog'}}
vis_utils.visualize_boxes_and_labels_on_image_array(
test_image,
detections[:, :4],
detections[:, 4].astype(np.int32),
detections[:, 5],
labelmap,
track_ids=None,
use_normalized_coordinates=True,
max_boxes_to_draw=1,
min_score_thresh=0.2,
agnostic_mode=False,
line_thickness=8)
self.assertGreater(np.abs(np.sum(test_image - ori_image)), 0)
def visualize(self):
"""save tfrecords images with bounding boxes."""
vis_ds = self.input_fn(params=self.params)
data = next(iter(vis_ds)) # iterable.
images = data[0]
gt_data = data[1]['groundtruth_data']
# scales
scale_to_org = data[1]['image_scales']
scales = 1.0 / scale_to_org
offset = tf.constant([0.485, 0.456, 0.406])
offset = tf.reshape(offset, (1, 1, -1))
scale_image = tf.constant([0.229, 0.224, 0.225])
scale_image = tf.reshape(scale_image, (1, 1, -1))
logging.info('Visualizing TfRecords %s', FLAGS.file_pattern)
for i, zip_data in enumerate(zip(gt_data, images, scales)):
gt, image, scale = zip_data
boxes = gt[:, :4]
boxes = boxes[np.any(boxes > 0, axis=1)].numpy()
if boxes.shape[0] > 0:
classes = gt[:boxes.shape[0], -1].numpy()
try:
category_index = {
idx: {
'id': idx,
'name': self.cls_to_label[idx]
}
for idx in np.asarray(classes, dtype=np.int)
}
except Exception: # pylint: disable=broad-except
category_index = {}
# unnormalize image.
image *= scale_image
image += offset
# 0-255. range
image = np.asarray(image.numpy() * 255., dtype=np.uint8)
# scale to image_size
boxes *= scale.numpy()
image = vis_utils.visualize_boxes_and_labels_on_image_array(
image,
boxes=boxes,
classes=classes.astype(int),
scores=np.ones(boxes.shape[0]),
category_index=category_index,
line_thickness=2,
skip_scores=True)
image = Image.fromarray(image)
image.save(
os.path.join(FLAGS.save_samples_dir, f'sample{i}.jpg'))
def visualize_image(image,
boxes,
classes,
scores,
label_map=None,
min_score_thresh=0.01,
max_boxes_to_draw=1000,
line_thickness=2,
**kwargs):
"""Visualizes a given image.
Args:
image: a image with shape [H, W, C].
boxes: a box prediction with shape [N, 4] ordered [ymin, xmin, ymax, xmax].
classes: a class prediction with shape [N].
scores: A list of float value with shape [N].
label_map: a dictionary from class id to name.
min_score_thresh: minimal score for showing. If claass probability is below
this threshold, then the object will not show up.
max_boxes_to_draw: maximum bounding box to draw.
line_thickness: how thick is the bounding box line.
**kwargs: extra parameters.
Returns:
output_image: an output image with annotated boxes and classes.
"""
label_map = label_util.get_label_map(label_map or 'coco')
category_index = {k: {'id': k, 'name': label_map[k]} for k in label_map}
img = np.array(image)
vis_utils.visualize_boxes_and_labels_on_image_array(
img,
boxes,
classes,
scores,
category_index,
min_score_thresh=min_score_thresh,
max_boxes_to_draw=max_boxes_to_draw,
line_thickness=line_thickness,
**kwargs)
return img
def visualize_image(image,
boxes,
classes,
scores,
id_mapping=None,
min_score_thresh=anchors.MIN_SCORE_THRESH,
max_boxes_to_draw=anchors.MAX_DETECTIONS_PER_IMAGE,
line_thickness=2,
**kwargs):
"""Visualizes a given image.
Args:
image: a image with shape [H, W, C].
boxes: a box prediction with shape [N, 4] ordered [ymin, xmin, ymax, xmax].
classes: a class prediction with shape [N].
scores: A list of float value with shape [N].
id_mapping: a dictionary from class id to name.
min_score_thresh: minimal score for showing. If claass probability is below
this threshold, then the object will not show up.
max_boxes_to_draw: maximum bounding box to draw.
line_thickness: how thick is the bounding box line.
**kwargs: extra parameters.
Returns:
output_image: an output image with annotated boxes and classes.
"""
id_mapping = parse_label_id_mapping(id_mapping)
category_index = {k: {'id': k, 'name': id_mapping[k]} for k in id_mapping}
img = np.array(image)
vis_utils.visualize_boxes_and_labels_on_image_array(
img,
boxes,
classes,
scores,
category_index,
min_score_thresh=min_score_thresh,
max_boxes_to_draw=max_boxes_to_draw,
line_thickness=line_thickness,
**kwargs)
return img
def visualize_image(image,
boxes,
classes,
scores,
id_mapping,
min_score_thresh=0.2,
max_boxes_to_draw=50,
line_thickness=4,
**kwargs):
category_index = {k: {'id': k, 'name': id_mapping[k]} for k in id_mapping}
img = np.array(image)
vis_utils.visualize_boxes_and_labels_on_image_array(
img,
boxes,
classes,
scores,
category_index,
min_score_thresh=min_score_thresh,
max_boxes_to_draw=max_boxes_to_draw,
line_thickness=line_thickness,
**kwargs)
return img
