def parse_image(cv_image):
candidate, subset = body_estimation(cv_image)
canvas = copy.deepcopy(cv_image)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, cv_image)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
peaks = hand_estimation(cv_image[y:y + w, x:x + w, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0], peaks[:, 0] + x)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1], peaks[:, 1] + y)
all_hand_peaks.append(peaks)
canvas = util.draw_handpose(canvas, all_hand_peaks)
rgb_image = canvas
skeleton = [(-1, -1)] * 18
for i in range(18):
for n in range(len(subset)):
index = int(subset[n][i])
if index == -1:
continue
x, y = candidate[index][0:2]
skeleton[i] = (x, y)
return rgb_image, skeleton, [peaks.tolist() for peaks in all_hand_peaks]
from hand import Hand
from body import Body
import matplotlib.pyplot as plt
import copy
import numpy as np
body_estimation = Body('model/body_pose_model.pth')
hand_estimation = Hand('model/hand_pose_model.pth')
test_image = 'images/demo.jpg'
oriImg = cv2.imread(test_image) # B,G,R order
candidate, subset = body_estimation(oriImg)
canvas = copy.deepcopy(oriImg)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
# cv2.rectangle(canvas, (x, y), (x+w, y+w), (0, 255, 0), 2, lineType=cv2.LINE_AA)
# cv2.putText(canvas, 'left' if is_left else 'right', (x, y), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
# if is_left:
# plt.imshow(oriImg[y:y+w, x:x+w, :][:, :, [2, 1, 0]])
# plt.show()
peaks = hand_estimation(oriImg[y:y + w, x:x + w, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0], peaks[:, 0] + x)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1], peaks[:, 1] + y)
# else:
# peaks = hand_estimation(cv2.flip(oriImg[y:y+w, x:x+w, :], 1))
# peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], w-peaks[:, 0]-1+x)
def detect(oriImg, image_name, mode, model, hand_estimation, body_estimation, rpn, labels):
height, width, channel = oriImg.shape
canvas = copy.deepcopy(oriImg)
if mode == "openpose":
candidate, subset = body_estimation(oriImg)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)
if rpn and labels:
# detect object
detections , scale = detect_object.detect(rpn, oriImg)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
cv2.rectangle(canvas, (x, y), (x+w, y+w), (0, 255, 0), 2, lineType=cv2.LINE_AA)
peaks = hand_estimation(oriImg[y:y+w, x:x+w, :])
peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], peaks[:, 0]+x)
peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
all_hand_peaks.append(peaks)
elif mode == "handpose":
bounding_box = inference_detector(model, oriImg)
if rpn and labels:
# detect object
detections , scale = detect_object.detect(rpn, oriImg)
all_hand_peaks = []
for xmin, ymin, xmax, ymax, prob in bounding_box[0]:
if prob < 0.4:
continue
fixed_xmin = int(xmin) - 50
if fixed_xmin <=0:
fixed_xmin = 1
fixed_xmax = int(xmax) + 50
if fixed_xmax >= width:
fixed_xmax = width - 1
fixed_ymin = int(ymin) - 50
if fixed_ymin <=0:
fixed_ymin = 1
fixed_ymax = int(ymax) + 50
if fixed_ymax >= height:
fixed_ymax = height - 1
cv2.rectangle(canvas, (fixed_xmin, fixed_ymin), (fixed_xmax, fixed_ymax), (0, 0, 255), 2)
peaks = hand_estimation(oriImg[fixed_ymin:fixed_ymax, fixed_xmin:fixed_xmax, :])
peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], peaks[:, 0]+fixed_xmin)
peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+fixed_ymin)
all_hand_peaks.append(peaks)
if rpn and labels:
canvas = detect_object.draw_ssd_bbox(canvas, detections, scale, labels)
if all_hand_peaks:
canvas = util.draw_handpose(canvas, all_hand_peaks)
canvas = cv2.resize(canvas,(width, height))
cv2.imwrite(os.path.join('demo','output',image_name),canvas)
return
def main(webcamera=False):
args = get_option()
if args.object_detection:
rpn = ssd_setting.build_ssd('test', 300, 21)
rpn.load_weights(args.ssd_weight_path)
from ssd.data import VOC_CLASSES as labels
#bounding box model
model = init_detector(args.config_file, args.weight_file, device="cuda:0")
hand_estimation = Hand(args.hand_weight_path)
if args.mode == "openpose":
body_estimation = Body(args.body_weight_path)
if args.video_name == "webcam":
video = cv2.VideoCapture(0)
else:
video = cv2.VideoCapture(os.path.join('demo', 'input',
args.video_name))
fps = int(video.get(cv2.CAP_PROP_FPS))
width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
fourcc = cv2.VideoWriter_fourcc("m", "p", "4", "v")
new_video = cv2.VideoWriter(
os.path.join('demo', 'output',
args.video_name.split('.')[0] + ".mp4"), fourcc, fps,
(width, height))
i = 0
start = time.perf_counter()
while True:
ret, frame = video.read()
if ret:
canvas = copy.deepcopy(frame)
if args.mode == "openpose":
candidate, subset = body_estimation(frame)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, frame)
# detect object
detections, scale = detect_object.detect(rpn, frame)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
cv2.rectangle(canvas, (x, y), (x + w, y + w), (0, 255, 0),
2,
lineType=cv2.LINE_AA)
peaks = hand_estimation(frame[y:y + w, x:x + w, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + x)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + y)
all_hand_peaks.append(peaks)
elif args.mode == "handpose":
bounding_box = inference_detector(model, frame)
# detect object
detections, scale = detect_object.detect(rpn, frame)
all_hand_peaks = []
for xmin, ymin, xmax, ymax, prob in bounding_box[0]:
if prob < 0.5:
continue
fixed_xmin = int(xmin) - 50
if fixed_xmin <= 0:
fixed_xmin = 1
fixed_xmax = int(xmax) + 50
if fixed_xmax >= width:
fixed_xmax = width - 1
fixed_ymin = int(ymin) - 50
if fixed_ymin <= 0:
fixed_ymin = 1
fixed_ymax = int(ymax) + 50
if fixed_ymax >= height:
fixed_ymax = height - 1
cv2.rectangle(canvas, (fixed_xmin, fixed_ymin),
(fixed_xmax, fixed_ymax), (0, 0, 255), 2)
peaks = hand_estimation(frame[fixed_ymin:fixed_ymax,
fixed_xmin:fixed_xmax, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + fixed_xmin)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + fixed_ymin)
all_hand_peaks.append(peaks)
canvas = detect_object.draw_ssd_bbox(canvas, detections, scale,
labels)
if all_hand_peaks:
canvas = util.draw_handpose(canvas, all_hand_peaks)
canvas = cv2.resize(canvas, (width, height))
new_video.write(canvas)
i = i + 1
print(i)
else:
break
end = time.perf_counter()
print(end - start)
new_video.release()
return
def main():
args = get_option()
image_name = args.image_name
config_file = args.config_file
weight_file = args.weight_file
#model load
model = init_detector(config_file, weight_file, device="cuda:0")
hand_estimation = Hand('model/hand_pose_model.pth')
test_image = (os.path.join('demo', 'input', image_name))
oriImg = cv2.imread(test_image) # B,G,R order
height, width, channel = oriImg.shape
canvas = copy.deepcopy(oriImg)
if args.mode == "openpose":
body_estimation = Body('model/body_pose_model.pth')
candidate, subset = body_estimation(oriImg)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
cv2.rectangle(canvas, (x, y), (x + w, y + w), (0, 255, 0),
2,
lineType=cv2.LINE_AA)
# cv2.putText(canvas, 'left' if is_left else 'right', (x, y), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
# if is_left:
# plt.imshow(oriImg[y:y+w, x:x+w, :][:, :, [2, 1, 0]])
# plt.show()
peaks = hand_estimation(oriImg[y:y + w, x:x + w, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + x)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + y)
# else:
# peaks = hand_estimation(cv2.flip(oriImg[y:y+w, x:x+w, :], 1))
# peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], w-peaks[:, 0]-1+x)
# peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
# print(peaks)
all_hand_peaks.append(peaks)
elif args.mode == "handpose":
bounding_box = inference_detector(model, oriImg)
all_hand_peaks = []
for xmin, ymin, xmax, ymax, prob in bounding_box[0]:
if prob < 0.4:
continue
fixed_xmin = int(xmin) - 50
if fixed_xmin <= 0:
fixed_xmin = 1
fixed_xmax = int(xmax) + 50
if fixed_xmax >= width:
fixed_xmax = width - 1
fixed_ymin = int(ymin) - 50
if fixed_ymin <= 0:
fixed_ymin = 1
fixed_ymax = int(ymax) + 50
if fixed_ymax >= height:
fixed_ymax = height - 1
cv2.rectangle(canvas, (fixed_xmin, fixed_ymin),
(fixed_xmax, fixed_ymax), (0, 0, 255), 2)
peaks = hand_estimation(oriImg[fixed_ymin:fixed_ymax,
fixed_xmin:fixed_xmax, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + fixed_xmin)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + fixed_ymin)
all_hand_peaks.append(peaks)
if all_hand_peaks:
canvas = util.draw_handpose(canvas, all_hand_peaks)
canvas = cv2.resize(canvas, (width, height))
cv2.imwrite(os.path.join('demo', 'output', image_name), canvas)
return
def main(webcamera=False):
args = get_option()
video_name = args.video_name
config_file = args.config_file
weight_file = args.weight_file
#bounding box model
model = init_detector(config_file, weight_file, device="cuda:0")
hand_estimation = Hand('model/hand_pose_model.pth')
body_estimation = Body('model/body_pose_model.pth')
if video_name == "webcam":
video = cv2.VideoCapture(0)
else:
video = cv2.VideoCapture(os.path.join('demo', 'input', video_name))
fps = int(video.get(cv2.CAP_PROP_FPS))
width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT))
fourcc = cv2.VideoWriter_fourcc("m", "p", "4", "v")
new_video = cv2.VideoWriter(os.path.join('demo', 'output', video_name),
fourcc, fps, (width, height))
i = 0
start = time.perf_counter()
while True:
ret, frame = video.read()
if ret:
canvas = copy.deepcopy(frame)
if args.mode == "openpose":
candidate, subset = body_estimation(frame)
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, frame)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
cv2.rectangle(canvas, (x, y), (x + w, y + w), (0, 255, 0),
2,
lineType=cv2.LINE_AA)
# cv2.putText(canvas, 'left' if is_left else 'right', (x, y), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
# if is_left:
# plt.imshow(oriImg[y:y+w, x:x+w, :][:, :, [2, 1, 0]])
# plt.show()
peaks = hand_estimation(frame[y:y + w, x:x + w, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + x)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + y)
# else:
# peaks = hand_estimation(cv2.flip(oriImg[y:y+w, x:x+w, :], 1))
# peaks[:, 0] = np.where(peaks[:, 0]==0, peaks[:, 0], w-peaks[:, 0]-1+x)
# peaks[:, 1] = np.where(peaks[:, 1]==0, peaks[:, 1], peaks[:, 1]+y)
# print(peaks)
all_hand_peaks.append(peaks)
elif args.mode == "handpose":
bounding_box = inference_detector(model, frame)
all_hand_peaks = []
for xmin, ymin, xmax, ymax, prob in bounding_box[0]:
if prob < 0.5:
continue
fixed_xmin = int(xmin) - 50
if fixed_xmin <= 0:
fixed_xmin = 1
fixed_xmax = int(xmax) + 50
if fixed_xmax >= width:
fixed_xmax = width - 1
fixed_ymin = int(ymin) - 50
if fixed_ymin <= 0:
fixed_ymin = 1
fixed_ymax = int(ymax) + 50
if fixed_ymax >= height:
fixed_ymax = height - 1
cv2.rectangle(canvas, (fixed_xmin, fixed_ymin),
(fixed_xmax, fixed_ymax), (0, 0, 255), 2)
peaks = hand_estimation(frame[fixed_ymin:fixed_ymax,
fixed_xmin:fixed_xmax, :])
peaks[:, 0] = np.where(peaks[:, 0] == 0, peaks[:, 0],
peaks[:, 0] + fixed_xmin)
peaks[:, 1] = np.where(peaks[:, 1] == 0, peaks[:, 1],
peaks[:, 1] + fixed_ymin)
all_hand_peaks.append(peaks)
if all_hand_peaks:
canvas = util.draw_handpose(canvas, all_hand_peaks)
canvas = cv2.resize(canvas, (width, height))
new_video.write(canvas)
i = i + 1
print(i)
else:
break
end = time.perf_counter()
print(end - start)
new_video.release()
return