def test_rgb2gray_gray2rgb():
im_rgb = mv.gray2rgb(IM_GRAY)
im_gray_2 = mv.rgb2gray(im_rgb)
assert_image_equal(IM_GRAY, im_gray_2)
im_gray_2 = cv2.cvtColor(im_rgb, cv2.COLOR_RGB2GRAY)
assert_image_equal(IM_GRAY, im_gray_2)
def normalize_rgb(img, mean, std):
""" Normalize an image.
Subtract mean per channel and divide by std per channel. (support
grayscale image and RGB image).
Args:
img (ndarray): image to be normalized.
mean (tuple[float] or float): mean values.
std (tuple[float] or float): standard deviations.
Return:
(ndarray): the normalized RGB image.
Note:
For grayscale image, first rescale intensity to [0.0, 255.0].
Then convert into a 3-channel RGB image before normalization.
"""
img = img.astype(np.float32)
if img.ndim == 2:
img = normalize_grayscale(img, to_float=False) * 255.0
img = mv.gray2rgb(img)
return (img - mean) / std
def imshow_bboxes(img,
bboxes,
score_thr=0,
colors=Color.green,
top_k=-1,
thickness=1,
font_scale=0.5,
show=True,
title='',
save_path=None):
""" Draw bounding boxes on an image.
To display detection result or compare detection results by different
algorithms.
Args:
img (str or ndarray): image (or file path) to be displayed.
bboxes (list or ndarray): a list of ndarray of shape (k, 4) or (n, 5).
score_thr (float): minimum score of bboxes to be shown.
colors (Color or list[Color]): color or list of colors.
top_k (int): plot the first k bboxes only if set positive.
Otherwise, plot all the bboxes.
thickness (int): line thickness.
font_scale (float): font scales of texts.
show (bool): True: show the image; False: save the image.
title (str): title of the plot.
save_path (str, optional): path to save the image.
"""
if isinstance(img, str):
img = mv.imread(img)
else:
img = img if img.ndim == 3 else mv.gray2rgb(img)
if isinstance(bboxes, np.ndarray):
bboxes = [bboxes]
if isinstance(colors, Color):
colors = [colors] * len(bboxes)
assert len(bboxes) == len(colors)
plot_prob = True if bboxes[0].shape[1] == 5 else False
if score_thr > 0:
for i in range(len(bboxes)):
assert bboxes[i].shape[1] == 5
scores = bboxes[i][:, -1]
indices = scores > score_thr
bboxes[i] = bboxes[i][indices, :]
img_with_result = img.copy()
for i, _bboxes in enumerate(bboxes):
_bboxes_int = _bboxes.astype(np.int32)
if top_k <= 0:
_top_k = _bboxes.shape[0]
else:
_top_k = min(top_k, _bboxes.shape[0])
for j in range(_top_k):
left_top = (_bboxes_int[j, 0], _bboxes_int[j, 1])
right_bottom = (_bboxes_int[j, 2], _bboxes_int[j, 3])
cv2.rectangle(img_with_result, left_top, right_bottom,
colors[i].value, thickness)
if plot_prob:
label_text = '%.2f' % _bboxes[j, -1]
cv2.putText(img_with_result, label_text,
(_bboxes_int[j, 0], _bboxes_int[j, 1] - 2),
cv2.FONT_HERSHEY_COMPLEX, font_scale,
colors[i].value, thickness)
if show:
_imshow_switcher([img_with_result, img], title)
if save_path is not None:
mv.imwrite(save_path, img_with_result)