def run(model, img):
img_show = img.copy()
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
h, w, c = img.shape
# 只在最右边或者最下边填充0, 这样不会影响box或者点的坐标值, 所以无需再对box或点的坐标做改变
if w > h:
img = cv2.copyMakeBorder(img, 0, w - h, 0, 0, cv2.BORDER_CONSTANT, value=(0, 0, 0))
else:
img = cv2.copyMakeBorder(img, 0, 0, 0, h - w, cv2.BORDER_CONSTANT, value=(0, 0, 0))
img_ori = img.copy()
img = cv2.resize(img, (netH, netW), interpolation=cv2.INTER_CUBIC)
img_input = np.expand_dims(img.astype(np.float32) / 255., axis=0)
factor_x = img_ori.shape[1] / (netW / 4)
factor_y = img_ori.shape[0] / (netH / 4)
center_map, kps_reg_map = infer(model, img_input)
# label = outputs[0]
spm_decoder = SpmDecoder(factor_x, factor_y, netH // 4, netW // 4)
joints, centers = spm_decoder([center_map[0], kps_reg_map[0]], score_thres=score, dis_thres=dist)
print(centers)
for j, single_person_joints in enumerate(joints):
cv2.circle(img_show, (int(centers[j][0]), int(centers[j][1])), 8, colors[j % 6], thickness=-1)
for i in range(14):
x = int(single_person_joints[2 * i])
y = int(single_person_joints[2 * i + 1])
cv2.circle(img_show, (x, y), 4, colors[j % 6], thickness=-1)
# cv2.putText(img_show, str(i), (x, y), cv2.FONT_HERSHEY_COMPLEX, 1, (0, 0, 250), 1)
cv2.imshow('result', img_show)
k = cv2.waitKey(0)
def run(model, img):
img_show = img.copy()
h, w, c = img.shape
# 只在最右边或者最下边填充0, 这样不会影响box或者点的坐标值, 所以无需再对box或点的坐标做改变
#if w > h:
# img = cv2.copyMakeBorder(img, 0, w - h, 0, 0, cv2.BORDER_CONSTANT, value=(0, 0, 0))
#else:
# img = cv2.copyMakeBorder(img, 0, 0, 0, h - w, cv2.BORDER_CONSTANT, value=(0, 0, 0))
img = img.astype(np.float32) / 255.0
img = cv2.resize(img, (netH, netW), interpolation=cv2.INTER_CUBIC)
img_input = np.expand_dims(img, axis=0)
factor_x = img_show.shape[1] / (netW / 4)
factor_y = img_show.shape[0] / (netH / 4)
center_map, kps_reg_map = infer(model, img_input)
# print (center_map.shape)
# print (kps_reg_map.shape)
# label = outputs[0]
spm_decoder = SpmDecoder(factor_x, factor_y, netH // 4, netW // 4)
results = spm_decoder([center_map[0].numpy(), kps_reg_map[0].numpy()],
score_thres=score,
dis_thres=dist)
for j, result in enumerate(results):
# print (result)
center = result['center']
single_person_joints = result['joints']
cv2.circle(img_show, (int(center[0]), int(center[1])),
5,
colors[j % 3],
thickness=-1)
for i in range(12):
x = int(single_person_joints[2 * i])
y = int(single_person_joints[2 * i + 1])
cv2.circle(img_show, (x, y), 4, colors[j % 3], thickness=-1)
cv2.putText(img, str(i), (x, y), cv2.FONT_HERSHEY_COMPLEX, 1,
(0, 0, 250), 1)
return img_show
cv2.BORDER_CONSTANT,
value=(0, 0, 0))
img_ori = img.copy()
img = cv2.resize(img, (netH, netW), interpolation=cv2.INTER_CUBIC)
img_input = np.expand_dims(img.astype(np.float32) / 255., axis=0)
factor_x = img_ori.shape[1] / (netW / 4)
factor_y = img_ori.shape[0] / (netH / 4)
# factor_x = 4
# factor_y = 4
# img_ori = img
center_map, kps_reg_map = model(img_input)
# label = outputs[0]
spm_decoder = SpmDecoder(factor_x, factor_y, netH // 4, netW // 4)
joints, centers = spm_decoder([center_map[0], kps_reg_map[0]],
score_thres=score,
dis_thres=dist)
print(centers)
for j, single_person_joints in enumerate(joints):
cv2.circle(img_ori, (int(centers[j][0]), int(centers[j][1])),
8,
colors[j % 6],
thickness=-1)
for i in range(14):
x = int(single_person_joints[2 * i])
y = int(single_person_joints[2 * i + 1])
cv2.circle(img_ori, (x, y), 4, colors[j % 6], thickness=-1)
cv2.putText(img_ori, str(i), (x, y), cv2.FONT_HERSHEY_COMPLEX,
mode = 'train'
dataset = get_dataset(mode=mode)
colors = [[0,0,255],[255,0,0],[0,255,0]]
for epco in range(1):
for step, (img, center_map, center_mask, kps_map, kps_map_weight) in enumerate(dataset):
# print (step)
# print (img[0].shape)
# img1 = img[0]
# label1 = label[0]
# break
print ('epoch {} / step {}'.format(epco, step))
img = (img.numpy()[0] * 255).astype(np.uint8)
spm_decoder = SpmDecoder(4, 4, params['height']//4, params['width']//4)
results = spm_decoder([center_map[0].numpy(), kps_map[0].numpy()])
for j, result in enumerate(results):
center = result['center']
single_person_joints = result['joints']
cv2.circle(img, (int(center[0]), int(center[1])), 5, colors[j%3], thickness=-1)
for i in range(params['num_joints']):
x = int(single_person_joints[2*i])
y = int(single_person_joints[2*i+1])
cv2.circle(img, (x,y), 4, colors[j%3],thickness=-1)
cv2.imshow('label', img)
k = cv2.waitKey(0)
if k == 113:
break
print(str(imgids[0], encoding='utf-8'))
break
else:
for epco in range(1):
for step, (img, label) in enumerate(dataset):
# print (step)
# print (img[0].shape)
# img1 = img[0]
# label1 = label[0]
# break
print('epoch {} / step {}'.format(epco, step))
img = (img.numpy()[0] * 255).astype(np.uint8)
label = label[0]
spm_decoder = SpmDecoder(4, 4, center_config['height'] // 4,
center_config['width'] // 4)
joints, centers = spm_decoder(label)
for j, single_person_joints in enumerate(joints):
cv2.circle(img, (int(centers[j][0]), int(centers[j][1])),
8,
colors[j % 3],
thickness=-1)
for i in range(14):
x = int(single_person_joints[2 * i])
y = int(single_person_joints[2 * i + 1])
cv2.circle(img, (x, y), 4, colors[j % 3], thickness=-1)
cv2.putText(img, str(i), (x, y),
cv2.FONT_HERSHEY_COMPLEX, 1, (0, 0, 250),
1)
cv2.imshow('label', img)