def estimate_all(folder, model):
if not os.path.isdir(folder):
print('Warning: Directory does not exist...')
return
pose_folder = folder.replace('images', 'poses')
if os.path.isdir(pose_folder):
print('Cached poses found')
else:
os.mkdir(pose_folder)
image_list = [
name for name in os.listdir(folder) if name.endswith('.jpg')
]
for name in tqdm(image_list):
im_name = os.path.join(folder, name)
img = imread(im_name)
pose_cords = np.array(cordinates_from_image_file(img, model=model))
new_path = im_name.replace(folder, pose_folder)
color, _ = pose_utils.draw_pose_from_cords(pose_cords, (256, 256))
imsave(new_path, color)
new_path = new_path.replace('.jpg', '.pose.npy')
np.save(new_path, pose_cords)
return pose_folder
df = pd.read_csv(args.annotations_file_test, sep=':')
ano_fr = df[df['name'] == fr].iloc[0]
ano_to = df[df['name'] == to].iloc[0]
kp_fr = pose_utils.load_pose_cords_from_strings(ano_fr['keypoints_y'],
ano_fr['keypoints_x'])
kp_to = pose_utils.load_pose_cords_from_strings(ano_to['keypoints_y'],
ano_to['keypoints_x'])
mask = pose_transform.pose_masks(kp_to,
img_size=args.image_size).astype(bool)
mask = np.array(reduce(np.logical_or, list(mask)))
mask = mask.astype('float')
pose_fr, _ = pose_utils.draw_pose_from_cords(kp_fr, args.image_size)
pose_to, _ = pose_utils.draw_pose_from_cords(kp_to, args.image_size)
cur_folder = os.path.join(out_folder, str(n))
if not os.path.exists(cur_folder):
os.makedirs(cur_folder)
imsave(os.path.join(cur_folder, 'from.jpg'),
imread(os.path.join(args.images_dir_test, fr)))
imsave(os.path.join(cur_folder, 'to.jpg'),
imread(os.path.join(args.images_dir_test, to)))
imsave(os.path.join(cur_folder, 'frpose.jpg'), pose_fr)
imsave(os.path.join(cur_folder, 'topose.jpg'), pose_to)
# imsave(os.path.join(cur_folder, 'mask.jpg'), mask)
cords[k] = _joint_num
return _joint_num, mask
if __name__ == "__main__":
img_dir = './datasets/cardio_dance_test/test_B' # Change this line into where your video frames are stored
pose_dir = img_dir.replace('test_B', 'test_A')
pose_npy_name = img_dir.replace('test_B', 'poses.npy')
if not os.path.isdir(pose_dir):
os.mkdir(pose_dir)
model = load_model('./pose_estimator/pose_estimator.h5')
img_list = os.listdir(img_dir)
# get frame shape
#print("list==", img_list)
tmp = imread(os.path.join(img_dir, img_list[0]))
im_shape = tmp.shape[:-1]
#print('tmp===', tmp)
#print('im_shape===', im_shape)
pose_cords = []
for item in tqdm(img_list):
img = imread(os.path.join(img_dir, item))
print("数组的维度数目", img.ndim)
cord = cordinates_from_image_file(img, model=model)
pose_cords.append(cord)
color, _ = draw_pose_from_cords(cord, im_shape)
imsave(os.path.join(pose_dir, item), color)
np.save(pose_npy_name, np.array(pose_cords, dtype=np.int))
plt.subplot(3, 1, 1)
kp_fr = pose_utils.load_pose_cords_from_strings(
kp_fr['keypoints_y'], kp_fr['keypoints_x'])
kp_to = pose_utils.load_pose_cords_from_strings(
kp_to['keypoints_y'], kp_to['keypoints_x'])
img = fr_img.copy()
#p, m = pose_utils.draw_pose_from_cords(kp_fr, img.shape[:2])
img = make_stickman(kp_to, fr_img.shape)
#img[m] = p[m]
plt.imshow(img)
plt.subplot(3, 1, 2)
img = to_img.copy()
p, m = pose_utils.draw_pose_from_cords(kp_to, img.shape[:2])
img[m] = p[m]
plt.imshow(img)
# tr = estimate_uniform_transform(kp_fr, kp_to)
#
# no_point_tr = np.array([[1, 0, 1000], [0, 1, 1000], [0, 0, 1]])
# if np.all(tr == no_point_tr.reshape((-1, 9))[..., :-1]):
# print >>f, '_'.join([fr,to]) + '.jpg'
p = resize(p, (256, 256), preserve_range=True).astype(np.uint8)
m = resize(m, (256, 256), preserve_range=True).astype(bool)
fr_img = resize(fr_img, (256, 256), preserve_range=True).astype(float)
tr = affine_transforms(kp_fr, kp_to)
masks = pose_masks(kp_to, fr_img.shape[:2])
