def classify_image(image_data):
img_array = np.fromstring(image_data, dtype=np.uint8)
image = cv.imdecode(img_array, cv.IMREAD_COLOR)
image_size = image.shape[:2]
x, y = random_choice(image_size)
image = safe_crop(image, x, y)
print('Start processing image...')
x_test = np.empty((1, img_rows, img_cols, 3), dtype=np.float32)
x_test[0, :, :, 0:3] = image / 255.
out = model.predict(x_test)
out = np.reshape(out, (img_rows, img_cols, num_classes))
out = np.argmax(out, axis=2)
out = to_bgr(out)
ret = image * 0.6 + out * 0.4
ret = ret.astype(np.uint8)
K.clear_session()
is_success, buffer = cv.imencode(".png", out)
buffer
out_bytes = BytesIO(buffer)
return out_bytes
def test_generate_trimap(self):
image = cv.imread('fg/1-1252426161dfXY.jpg')
alpha = cv.imread('mask/1-1252426161dfXY.jpg', 0)
trimap = generate_trimap(alpha)
self.assertEqual(trimap.shape, (615, 410))
# ensure np.where works as expected.
count = 0
h, w = trimap.shape[:2]
for i in range(h):
for j in range(w):
if trimap[i, j] == unknown_code:
count += 1
x_indices, y_indices = np.where(trimap == unknown_code)
num_unknowns = len(x_indices)
self.assertEqual(count, num_unknowns)
# ensure an unknown pixel is chosen
ix = random.choice(range(num_unknowns))
center_x = x_indices[ix]
center_y = y_indices[ix]
self.assertEqual(trimap[center_x, center_y], unknown_code)
x, y = random_choice(trimap)
# print(x, y)
image = safe_crop(image, x, y)
trimap = safe_crop(trimap, x, y)
alpha = safe_crop(alpha, x, y)
cv.imwrite('temp/test_generate_trimap_image.png', image)
cv.imwrite('temp/test_generate_trimap_trimap.png', trimap)
cv.imwrite('temp/test_generate_trimap_alpha.png', alpha)
def test_safe_crop(self):
name = 'SUNRGBD/kv1/NYUdata/NYU0899'
image, image_size = get_image(name)
semantic = get_category(name, image_size)
different_sizes = [(320, 320), (480, 480), (640, 640)]
crop_size = random.choice(different_sizes)
x, y = random_choice(image_size, crop_size)
image = safe_crop(image, x, y, crop_size)
semantic = safe_crop(semantic, x, y, crop_size)
semantic = to_bgr(semantic)
cv.imwrite('temp/test_safe_crop_image.png', image)
cv.imwrite('temp/test_safe_crop_semantic.png', semantic)
def test_flip(self):
image = cv.imread('fg/1-1252426161dfXY.jpg')
# print(image.shape)
alpha = cv.imread('mask/1-1252426161dfXY.jpg', 0)
trimap = generate_trimap(alpha)
x, y = random_choice(trimap)
image = safe_crop(image, x, y)
trimap = safe_crop(trimap, x, y)
alpha = safe_crop(alpha, x, y)
image = np.fliplr(image)
trimap = np.fliplr(trimap)
alpha = np.fliplr(alpha)
cv.imwrite('temp/test_flip_image.png', image)
cv.imwrite('temp/test_flip_trimap.png', trimap)
cv.imwrite('temp/test_flip_alpha.png', alpha)
def test_resize(self):
name = '0_0.png'
filename = os.path.join('merged', name)
image = cv.imread(filename)
bg_h, bg_w = image.shape[:2]
a = get_alpha(name)
a_h, a_w = a.shape[:2]
alpha = np.zeros((bg_h, bg_w), np.float32)
alpha[0:a_h, 0:a_w] = a
trimap = generate_trimap(alpha)
# 剪切尺寸 320:640:480 = 3:1:1
crop_size = (480, 480)
x, y = random_choice(trimap, crop_size)
image = safe_crop(image, x, y, crop_size)
trimap = safe_crop(trimap, x, y, crop_size)
alpha = safe_crop(alpha, x, y, crop_size)
cv.imwrite('temp/test_resize_image.png', image)
cv.imwrite('temp/test_resize_trimap.png', trimap)
cv.imwrite('temp/test_resize_alpha.png', alpha)
'''
filename = './single_human_paring/img/man/unexpose/img_023.jpg'
img_rows, img_cols = 320, 320
channel = 3
model_weights_path = 'Look_Into_Person/models/model.11-0.8409.hdf5'
model = build_model()
model.load_weights(model_weights_path)
print(model.summary())
image = cv.imread(filename)
image = cv.resize(image, (img_rows, img_cols), cv.INTER_CUBIC)
image_size = image.shape[:2]
x, y = random_choice(image_size)
image = safe_crop(image, x, y)
print('Start processing image: {}'.format(filename))
x_test = np.empty((1, img_rows, img_cols, 3), dtype=np.float32)
x_test[0, :, :, 0:3] = image / 255.
out = model.predict(x_test)
out = np.reshape(out, (img_rows, img_cols, num_classes))
out = np.argmax(out, axis=2)
out = to_bgr(out)
#
# ret = image * 0.6 + out * 0.4
# ret = ret.astype(np.uint8)
#if not os.path.exists('images'):
print('\nStart processing image: {}'.format(filename))
bgr_img = cv.imread(os.path.join(out_test_path, filename))
bg_h, bg_w = bgr_img.shape[:2]
print('bg_h, bg_w: ' + str((bg_h, bg_w)))
a = get_alpha_test(image_name)
a_h, a_w = a.shape[:2]
print('a_h, a_w: ' + str((a_h, a_w)))
alpha = np.zeros((bg_h, bg_w), np.float32)
alpha[0:a_h, 0:a_w] = a
trimap = generate_trimap(alpha)
different_sizes = [(320, 320), (320, 320), (320, 320), (480, 480), (640, 640)]
crop_size = random.choice(different_sizes)
x, y = random_choice(trimap, crop_size)
print('x, y: ' + str((x, y)))
bgr_img = safe_crop(bgr_img, x, y, crop_size)
alpha = safe_crop(alpha, x, y, crop_size)
trimap = safe_crop(trimap, x, y, crop_size)
cv.imwrite('images/{}_image.png'.format(i), np.array(bgr_img).astype(np.uint8))
cv.imwrite('images/{}_trimap.png'.format(i), np.array(trimap).astype(np.uint8))
cv.imwrite('images/{}_alpha.png'.format(i), np.array(alpha).astype(np.uint8))
x_test = np.empty((1, img_rows, img_cols, 4), dtype=np.float32)
x_test[0, :, :, 0:3] = bgr_img / 255.
x_test[0, :, :, 3] = trimap / 255.
y_true = np.empty((1, img_rows, img_cols, 2), dtype=np.float32)
y_true[0, :, :, 0] = alpha / 255.
filename = samples[i]
image_name = filename.split('.')[0]
print('Start processing image: {}'.format(filename))
x_test = np.empty((1, img_rows, img_cols, 4), dtype=np.float32)
bgr_img = cv.imread(os.path.join(out_test_path, filename))
bg_h, bg_w = bgr_img.shape[:2]
print(bg_h, bg_w)
a = get_alpha_test(image_name)
a_h, a_w = a.shape[:2]
print(a_h, a_w)
alpha = np.zeros((bg_h, bg_w), np.float32)
alpha[0:a_h, 0:a_w] = a
trimap = generate_trimap(alpha)
different_sizes = [(320, 320), (320, 320), (320, 320), (480, 480), (640, 640)]
crop_size = random.choice(different_sizes)
x, y = random_choice(trimap, crop_size)
print(x, y)
bgr_img = safe_crop(bgr_img, x, y, crop_size)
alpha = safe_crop(alpha, x, y, crop_size)
trimap = safe_crop(trimap, x, y, crop_size)
cv.imwrite('images/{}_image.png'.format(i), np.array(bgr_img).astype(np.uint8))
cv.imwrite('images/{}_trimap.png'.format(i), np.array(trimap).astype(np.uint8))
cv.imwrite('images/{}_alpha.png'.format(i), np.array(alpha).astype(np.uint8))
x_test = np.empty((1, 320, 320, 4), dtype=np.float32)
x_test[0, :, :, 0:3] = bgr_img / 255.
x_test[0, :, :, 3] = trimap / 255.
y_true = np.empty((1, 320, 320, 2), dtype=np.float32)
y_true[0, :, :, 0] = alpha / 255.
y_true[0, :, :, 1] = trimap / 255.
test_images = [f for f in os.listdir(rgb_test_path) if
os.path.isfile(os.path.join(rgb_test_path, f)) and f.endswith('.png')]
samples = random.sample(test_images, 10)
for i in range(len(samples)):
image_name = samples[i]
filename = os.path.join(rgb_test_path, image_name)
image = cv.imread(filename)
label = get_semantic(image_name)
image_size = image.shape[:2]
different_sizes = [(320, 320), (480, 480), (480, 480), (480, 480), (640, 640), (640, 640), (640, 640),
(960, 960), (960, 960), (960, 960)]
crop_size = random.choice(different_sizes)
x, y = random_choice(image_size, crop_size)
image = safe_crop(image, x, y, crop_size)
label = safe_crop(label, x, y, crop_size)
print('Start processing image: {}'.format(filename))
x_test = np.empty((1, img_rows, img_cols, 3), dtype=np.float32)
x_test[0, :, :, 0:3] = image / 255.
out = model.predict(x_test)
out = np.reshape(out, (img_rows, img_cols, num_classes))
out = np.argmax(out, axis=2)
out = to_bgr(out)
str_msg = 'crop_size: %s' % (str(crop_size))
draw_str(out, (20, 20), str_msg)