def Preprocessing(d, stage='train'):
height, width = cfg.data_shape
imgs = []
labels = []
valids = []
#asta e setat pe false in cfg
if cfg.use_seg:
segms = []
vis = False
#nu stiu cand se seteaza proprietatea de 'data', s-ar putea sa fie salvata direct din dataset. oricum, din ce arata in comentariu, va fi pathul la o imagine
img = d['data'] #cv2.imread(os.path.join(cfg.img_path, d['imgpath']))
# hack(multiprocessing data provider)
#daca nu e setata proprietatea 'data', cum s-ar putea sa nu fie
while img is None:
import pdb
pdb.set_trace()
print('read none image')
time.sleep(np.random.rand() * 5)
#img ia valoarea imaginii, ca practic asta se face cv2.imread
img = cv2.imread(os.path.join(cfg.img_path, d['imgpath']))
#shape[0] e height, shape[1] e width si shape[2] e channels
add = max(img.shape[0], img.shape[1])
#bimg e o imagine ca si img, doar ca la care adauga bordere de dimensiunea cea mai mare a imaginii, si de ceva culoare
bimg = cv2.copyMakeBorder(img,
add,
add,
add,
add,
borderType=cv2.BORDER_CONSTANT,
value=cfg.pixel_means.reshape(-1))
bbox = np.array(d['bbox']).reshape(4, ).astype(np.float32)
#la primele 2 coordonate ale bbox-ului, se adauga cea mai mare dimensiune a imaginii
bbox[:2] += add
if 'joints' in d:
#joints o sa fie o matrice cu 17 linii si 3 coloane. banuiesc ca din cauza ca sunt 17 jointuri luate in considerare
joints = np.array(d['joints']).reshape(cfg.nr_skeleton,
3).astype(np.float32)
#la primele 2 coloane se adauga cea mai mare dimensiune a imaginii
joints[:, :2] += add
#nu stiu de ce joints au si 3 coloane, dar daca a treia coloana are vreo valoare pe 0, se pun si celelalte 2 linii de la coloana respectiva pe -10000
#am aflat de ce sunt 3 coloane. primele 2 sunt x si y, si a treia banuiesc ca e z sau ceva depth
inds = np.where(joints[:, -1] == 0)
joints[inds, :2] = -1000000
#ideea e ca aici seteaza dimensiunile la care sa fie cropuita poza
crop_width = bbox[2] * (1 + cfg.imgExtXBorder * 2)
crop_height = bbox[3] * (1 + cfg.imgExtYBorder * 2)
#in asta e salvat centrul imaginii
objcenter = np.array([bbox[0] + bbox[2] / 2., bbox[1] + bbox[3] / 2.])
if stage == 'train':
crop_width = crop_width * (1 + 0.25)
crop_height = crop_height * (1 + 0.25)
#seteaza dimensiunea cea mai mica
if crop_height / height > crop_width / width:
crop_size = crop_height
min_shape = height
else:
crop_size = crop_width
min_shape = width
crop_size = min(crop_size, objcenter[0] / width * min_shape * 2. - 1.)
crop_size = min(
crop_size, (bimg.shape[1] - objcenter[0]) / width * min_shape * 2. - 1)
crop_size = min(crop_size, objcenter[1] / height * min_shape * 2. - 1.)
crop_size = min(crop_size,
(bimg.shape[0] - objcenter[1]) / height * min_shape * 2. -
1)
min_x = int(objcenter[0] - crop_size / 2. / min_shape * width)
max_x = int(objcenter[0] + crop_size / 2. / min_shape * width)
min_y = int(objcenter[1] - crop_size / 2. / min_shape * height)
max_y = int(objcenter[1] + crop_size / 2. / min_shape * height)
x_ratio = float(width) / (max_x - min_x)
y_ratio = float(height) / (max_y - min_y)
#asta e un soi de normalizare la dimensiunile dorite, si practic asigneaza valorile pt jointuri pt training si validare
if 'joints' in d:
joints[:, 0] = joints[:, 0] - min_x
joints[:, 1] = joints[:, 1] - min_y
joints[:, 0] *= x_ratio
joints[:, 1] *= y_ratio
label = joints[:, :2].copy()
valid = joints[:, 2].copy()
#imaginea devine imaginea cu border, dar cu widht si height
img = cv2.resize(bimg[min_y:max_y, min_x:max_x, :], (width, height))
if stage != 'train':
#un aray cu coordonatele care sunt diferenta dintre minimurile obtinute mai sus si cea mai mare dimensiune a imaginii initiale
details = np.asarray(
[min_x - add, min_y - add, max_x - add, max_y - add])
if cfg.use_seg is True and 'segmentation' in d:
seg = get_seg(ori_img.shape[0], ori_img.shape[1], d['segmentation'])
add = max(seg.shape[0], seg.shape[1])
bimg = cv2.copyMakeBorder(seg,
add,
add,
add,
add,
borderType=cv2.BORDER_CONSTANT,
value=(0, 0, 0))
seg = cv2.resize(bimg[min_y:max_y, min_x:max_x], (width, height))
segms.append(seg)
# e pus in functia asta pe false
if vis:
tmpimg = img.copy()
from utils.visualize import draw_skeleton
draw_skeleton(tmpimg, label.astype(int))
cv2.imwrite('vis.jpg', tmpimg)
from IPython import embed
embed()
#creadeam ca in pixel means e culoarea borderului da aparent sunt si ceva dimensiuni
img = img - cfg.pixel_means
if cfg.pixel_norm:
img = img / 255.
img = img.transpose(2, 0, 1)
imgs.append(img)
if 'joints' in d:
labels.append(label.reshape(-1))
valids.append(valid.reshape(-1))
if stage == 'train':
imgs, labels, valids = data_augmentation(imgs, labels, valids)
heatmaps15 = joints_heatmap_gen(imgs,
labels,
cfg.output_shape,
cfg.data_shape,
return_valid=False,
gaussian_kernel=cfg.gk15)
heatmaps11 = joints_heatmap_gen(imgs,
labels,
cfg.output_shape,
cfg.data_shape,
return_valid=False,
gaussian_kernel=cfg.gk11)
heatmaps9 = joints_heatmap_gen(imgs,
labels,
cfg.output_shape,
cfg.data_shape,
return_valid=False,
gaussian_kernel=cfg.gk9)
heatmaps7 = joints_heatmap_gen(imgs,
labels,
cfg.output_shape,
cfg.data_shape,
return_valid=False,
gaussian_kernel=cfg.gk7)
return [
imgs.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps15.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps11.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps9.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps7.astype(np.float32).transpose(0, 2, 3, 1),
valids.astype(np.float32)
]
else:
#practic intoarce un array cu imaginile normalizate si cu border, si un array cu dimensiunile
return [np.asarray(imgs).astype(np.float32), details]
def Preprocessing(d, stage='train'):
height, width = cfg.data_shape
imgs = []
labels = []
valids = []
if cfg.use_seg:
segms = []
vis = False
img = d['data']#cv2.imread(os.path.join(cfg.img_path, d['imgpath']))
# hack(multiprocessing data provider)
while img is None:
import pdb
pdb.set_trace()
print('read none image')
time.sleep(np.random.rand() * 5)
img = cv2.imread(os.path.join(cfg.img_path, d['imgpath']))
add = max(img.shape[0], img.shape[1])
bimg = cv2.copyMakeBorder(img, add, add, add, add, borderType=cv2.BORDER_CONSTANT,
value=cfg.pixel_means.reshape(-1))
bbox = np.array(d['bbox']).reshape(4, ).astype(np.float32)
bbox[:2] += add
if 'joints' in d:
joints = np.array(d['joints']).reshape(cfg.nr_skeleton, 3).astype(np.float32)
joints[:, :2] += add
inds = np.where(joints[:, -1] == 0)
joints[inds, :2] = -1000000
crop_width = bbox[2] * (1 + cfg.imgExtXBorder * 2)
crop_height = bbox[3] * (1 + cfg.imgExtYBorder * 2)
objcenter = np.array([bbox[0] + bbox[2] / 2., bbox[1] + bbox[3] / 2.])
if stage == 'train':
crop_width = crop_width * (1 + 0.25)
crop_height = crop_height * (1 + 0.25)
if crop_height / height > crop_width / width:
crop_size = crop_height
min_shape = height
else:
crop_size = crop_width
min_shape = width
crop_size = min(crop_size, objcenter[0] / width * min_shape * 2. - 1.)
crop_size = min(crop_size, (bimg.shape[1] - objcenter[0]) / width * min_shape * 2. - 1)
crop_size = min(crop_size, objcenter[1] / height * min_shape * 2. - 1.)
crop_size = min(crop_size, (bimg.shape[0] - objcenter[1]) / height * min_shape * 2. - 1)
min_x = int(objcenter[0] - crop_size / 2. / min_shape * width)
max_x = int(objcenter[0] + crop_size / 2. / min_shape * width)
min_y = int(objcenter[1] - crop_size / 2. / min_shape * height)
max_y = int(objcenter[1] + crop_size / 2. / min_shape * height)
x_ratio = float(width) / (max_x - min_x)
y_ratio = float(height) / (max_y - min_y)
if 'joints' in d:
joints[:, 0] = joints[:, 0] - min_x
joints[:, 1] = joints[:, 1] - min_y
joints[:, 0] *= x_ratio
joints[:, 1] *= y_ratio
label = joints[:, :2].copy()
valid = joints[:, 2].copy()
img = cv2.resize(bimg[min_y:max_y, min_x:max_x, :], (width, height))
if stage != 'train':
details = np.asarray([min_x - add, min_y - add, max_x - add, max_y - add])
if cfg.use_seg is True and 'segmentation' in d:
seg = get_seg(ori_img.shape[0], ori_img.shape[1], d['segmentation'])
add = max(seg.shape[0], seg.shape[1])
bimg = cv2.copyMakeBorder(seg, add, add, add, add, borderType=cv2.BORDER_CONSTANT, value=(0, 0, 0))
seg = cv2.resize(bimg[min_y:max_y, min_x:max_x], (width, height))
segms.append(seg)
if vis:
tmpimg = img.copy()
from utils.visualize import draw_skeleton
draw_skeleton(tmpimg, label.astype(int))
cv2.imwrite('vis.jpg', tmpimg)
from IPython import embed; embed()
img = img - cfg.pixel_means
if cfg.pixel_norm:
img = img / 255.
img = img.transpose(2, 0, 1)
imgs.append(img)
if 'joints' in d:
labels.append(label.reshape(-1))
valids.append(valid.reshape(-1))
if stage == 'train':
imgs, labels, valids = data_augmentation(imgs, labels, valids)
heatmaps15 = joints_heatmap_gen(imgs, labels, cfg.output_shape, cfg.data_shape, return_valid=False,
gaussian_kernel=cfg.gk15)
heatmaps11 = joints_heatmap_gen(imgs, labels, cfg.output_shape, cfg.data_shape, return_valid=False,
gaussian_kernel=cfg.gk11)
heatmaps9 = joints_heatmap_gen(imgs, labels, cfg.output_shape, cfg.data_shape, return_valid=False,
gaussian_kernel=cfg.gk9)
heatmaps7 = joints_heatmap_gen(imgs, labels, cfg.output_shape, cfg.data_shape, return_valid=False,
gaussian_kernel=cfg.gk7)
return [imgs.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps15.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps11.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps9.astype(np.float32).transpose(0, 2, 3, 1),
heatmaps7.astype(np.float32).transpose(0, 2, 3, 1),
valids.astype(np.float32)]
else:
return [np.asarray(imgs).astype(np.float32), details]