def video_to_images(videos, inDir, outDir):
for video in videos:
new_directory = os.path.join(outDir, os.path.splitext(video)[0])
if not os.path.isdir(new_directory):
os.mkdir(new_directory)
videocap = cv.VideoCapture(os.path.join(inDir, video))
success, image = videocap.read()
count = 0
while success:
#code.interact(local=dict(globals(), **locals()))
if "frame{}.jpg".format(count) in os.listdir(new_directory) and "frame{}.json".format(count) in os.listdir(new_directory):
print("frame{}.jpg and frame{}.json already in directory".format(count, count))
success, image = videocap.read()
count += 1
continue
cv.imwrite(os.path.join(new_directory, "frame{:05d}.jpg".format(count)), image)
print("{}: Saved frame {}".format(video, count))
datum = op.Datum()
datum.cvInputData = image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
json_file = os.path.join(new_directory, "frame{:05d}.json".format(count))
with open(json_file, "w") as f:
try:
json.dump(datum.poseKeypoints.tolist(), f)
except:
pass
success, image = videocap.read()
count += 1
def getOpenposeDataFrom(self, frame=None):
if (frame is None):
raise Exception("NoneType passed into OP wrapper class")
datum = op.Datum()
datum.cvInputData = frame
self.opWrapper.emplaceAndPop(op.VectorDatum([datum]))
return datum
def get_sample_heatmaps():
# These parameters are globally set. You need to unset variables set here if you have a new OpenPose object. See *
params = dict()
params["model_folder"] = "../../../models/"
params["heatmaps_add_parts"] = True
params["heatmaps_add_bkg"] = True
params["heatmaps_add_PAFs"] = True
params["heatmaps_scale"] = 3
params["upsampling_ratio"] = 1
params["body"] = 1
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image and get heatmap
datum = op.Datum()
imageToProcess = cv2.imread(args[0].image_path)
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
poseHeatMaps = datum.poseHeatMaps.copy()
opWrapper.stop()
return poseHeatMaps
def generate_keypoints(infile, show):
datum = op.Datum()
imageToProcess = cv2.imread(infile)
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# print("Body keypoints: \n" + str(datum.poseKeypoints))
print('Body keypoints shape:', datum.poseKeypoints.shape)
# print('Image shape:', datum.cvOutputData.shape)
image = datum.cvOutputData
data = {}
# show the keypoints of the FIRST person only
# for x, y, score in datum.poseKeypoints[0]:
# if score != 0:
# image = cv2.circle(image, (int(x), int(y)), radius=5, color=(0, 255, 255), thickness=-1)
if show:
cv2.imshow("OpenPose 1.7.0 - Tutorial Python API", image)
cv2.waitKey(0)
cv2.destroyAllWindows()
else:
outfile_data, outfile_pix = generate_outfile(infile=infile)
data['keypoints'] = datum.poseKeypoints
data['dimension'] = datum.cvOutputData.shape
np.save(outfile_data, data)
cv2.imwrite(outfile_pix, image)
print('Saving data to', outfile_data)
print('Saving pix to', outfile_pix)
def posemodl(image):
datum = op.Datum()
# imageToProcess = cv2.imread(args[0].image_path)
datum.cvInputData = image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# Display Image
# print("Body keypoints: \n" + str(datum.poseKeypoints))
# cv2.imshow("OpenPose 1.7.0 - Tutorial Python API", datum.cvOutputData)
# cv2.waitKey(0)
kp = datum.poseKeypoints
return kp, datum.cvOutputData
def body_from_image(self, image: np.ndarray) -> op.Datum:
"""
画像から骨格を推定します.
:param image: 画像
:return: 推定データ
"""
# 推定
datum = op.Datum()
datum.cvInputData = image
self.op_wrapper.emplaceAndPop(op.VectorDatum([datum]))
return datum
def opFrameRun(frame, opWrapper):
"""Applique openpose sur une image et retourne l'object datum contenant :
- l'image avec le squelette
- les keypoints en numpy
"""
datum = op.Datum()
datum.cvInputData = frame
#old version of pyopenpose library
#opWrapper.emplaceAndPop([datum])
#new version of pyopenpose library
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
return datum
def get_body_points(image_path):
try:
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath('__file__'))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../../python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + dir_path + '/../../../x64/Release;' + dir_path + '/../../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../../python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "../../../../models/"
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
datum = op.Datum()
imageToProcess = cv2.imread(image_path)
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# Display Image
print("Body keypoints: \n" + str(datum.poseKeypoints))
# cv2.imshow("OpenPose 1.6.0 - Tutorial Python API", datum.cvOutputData)
# cv2.waitKey(0)
return datum.poseKeypoints
except Exception as e:
print(e)
sys.exit(-1)
# body_points("C:/Users/lab/Desktop/taskey/action_video_test/2.jpg")
def getkey(img): #获取关键节点坐标
#旋转90度手机录像需要
dst_im = cv2.flip(
img, 0
) #原型：cv2.flip(src, flipCode[, dst]) → dst flipCode表示对称轴 0：x轴 1：y轴. -1：both
dst_im = cv2.transpose(dst_im)
im = cv2.resize(dst_im, (360, 640)) #设置为固定分辨率
datum = op.Datum()
#cv2.imshow('img',img)
#cv2.imshow('dst_im',dst_im)
#cv2.imshow('im',im)
#cv2.waitKey(0)
#cv2.destroyAllWindows()
datum.cvInputData = im
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
return datum.cvOutputData, datum.poseKeypoints #返回值1为骨架图，返回值2为坐标
def create_clips(file_name=None):
# Starting OpenPose
params = dict()
params["model_folder"] = "/home/rkaufman/dev/openpose/models"
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
out_dir = "/home/rkaufman/Downloads/restaurant_footage_openposified"
#while True:
# file_name = input("Next clip please.\n\t-->")
for file_name in clip_paths[3:]:
#while(True):
# file_name=input("next video please:\n\t-->")
cap = cv2.VideoCapture(file_name)
base, tail = os.path.split(file_name)
codec = cv2.VideoWriter_fourcc('M', 'J', 'P', 'G')
outfile_name = os.path.join(
out_dir, tail[:tail.find('.')] + "_openpose" + ".avi")
input_fps = 30
# input_fps = cap.get(cv2.CAP_PROP_FPS)
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
print("Writing to {}".format(outfile_name))
out_vid = cv2.VideoWriter(outfile_name, codec, input_fps,
(width, height))
i = 0
while (cap.isOpened()):
ret, frame = cap.read()
print("\titeration: {}".format(i))
if (not ret): break
datum = op.Datum()
datum.cvInputData = frame
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
op_image = datum.cvOutputData
# gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
out_vid.write(op_image)
#cv2.imshow('frame', frame)
#cv2.imshow('openpose', op_image)
i += 1
# if cv2.waitKey(1) & 0xFF == ord('q'):
# break
print("Successfully wrote to {}\n".format(outfile_name))
cap.release()
out_vid.release()
def get_skeleton(bbox,frame, player_index, offset=15):
x1,y1,x2,y2 = bbox.astype(int)
height, width, _ = frame.shape
# extract skeleton from bounding box
if y1-offset>=0 and x1-offset>=0 and y2+offset<height and x2+offset<width:
frame_ = frame[y1-offset:(y2)+offset,x1-offset:x2+offset, :]
else:
frame_ = frame[y1:y2,x1:x2, :]
scale_percent = 8
width = int(frame_.shape[1] * scale_percent / 10)
height = int(frame_.shape[0] * scale_percent / 10)
dim = (width, height)
src_img = cv2.resize(frame_, dim, interpolation=cv2.INTER_AREA)
datum.cvInputData = src_img
#print(datum.cvOutputData.shape)
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
#if datum.poseKeypoints is not None: print( datum.poseKeypoints[0,:,:])
if datum.poseKeypoints is not None:
cv2.imwrite("./results/test" + str(player_index) +"_"+ str("ss") + ".jpg",datum.cvOutputData)
keypoints = datum.poseKeypoints[0,:,:].tolist()
return keypoints
key = curr_item.replace('-', '')
if key not in params: params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = next_item
# Construct it from system arguments
# op.init_argv(args[1])
# oppython = op.OpenposePython()
# Starting OpenPose
opWrapper = op.WrapperPython(op.ThreadManagerMode.AsynchronousOut)
opWrapper.configure(params)
opWrapper.start()
# Main loop
userWantsToExit = False
while not userWantsToExit:
# Pop frame
datumProcessed = op.VectorDatum()
if opWrapper.waitAndPop(datumProcessed):
if not args[0].no_display:
# Display image
userWantsToExit = display(datumProcessed)
printKeypoints(datumProcessed)
else:
break
except Exception as e:
print(e)
sys.exit(-1)
def run_op(self, frame):
self.datum.cvInputData = frame
self.opWrp.emplaceAndPop(op.VectorDatum([self.datum]))
return self.datum.cvOutputData
def upload_file():
if request.method == 'POST':
# if 'file' not in request.files:
# flash('No file part')
# return redirect(request.url)
#
print(request.files.keys())
print(request.files['data'])
file = request.files['data']
print(dir(file))
print(file.filename)
file.save(os.path.join(os.getcwd(), 'uploads.mp4'))
try:
try:
import pyopenpose as op
except ImportError as e:
raise e
try:
import cv2
except ImportError as e:
raise e
try:
from test_video import SampleVideo
except ImportError as e:
raise e
params = dict()
params["model_folder"] = "C:\\openpose\\models\\"
params["number_people_max"] = 1
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
datum = op.Datum()
video_path = os.path.join(os.getcwd(), 'uploads.mp4')
cap = cv2.VideoCapture(video_path)
print(cap.get(cv2.CAP_PROP_FRAME_COUNT))
out_path = os.path.join(os.getcwd(), 'static/output.mp4')
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter(out_path, fourcc, 20.0,
(int(cap.get(3)), int(cap.get(4))))
for pos in range(int(cap.get(cv2.CAP_PROP_FRAME_COUNT))):
ret, frame = cap.read()
datum.cvInputData = frame
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
out.write(datum.cvOutputData)
print(datum.poseKeypoints)
print(cap.get(cv2.CAP_PROP_POS_FRAMES))
cap.release()
print('end')
except Exception as e:
print(dir(e))
print(e)
sys.exit(-1)
# if file.filename == '':
# flash('No selected file')
# return redirect(request.url)
# if file and allowed_file(file.filename):
# filename = secure_filename(file.filename)
# file.save(os.path.join(app.config['UPLOAD_FOLDER'], filename))
# return redirect(url_for('upload_file',
# filename=filename))
# return '''
# <!doctype html>
# <title>Upload new File</title>
# <h1>Upload new File</h1>
# <form method=post enctype=multipart/form-data>
# <input type=file name=file>
# <input type=submit value=Upload>
# </form>
# '''
return "good"
if request.method == 'GET':
return '''
def train(trainloader, validationloader):
### Initialization
params = list(cnn2.parameters()) + list(cnn1.parameters())
optimizer = torch.optim.Adam(params, lr=0.0001)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[100, 150], gamma=0.1)
dict1 = {'loss': [], 'valLoss': [], 'valPSNR': [], 'epoch': -1}
start = time.time()
cLoss = dict1['loss']
valLoss = dict1['valLoss']
valPSNR = dict1['valPSNR']
checkpoint_counter = 0
for epoch in range(dict1['epoch'] + 1, 3):
print("Epoch: ", epoch)
# Append and reset
cLoss.append([])
valLoss.append([])
valPSNR.append([])
iLoss = 0
# Increment scheduler count
scheduler.step()
for trainIndex, (trainData, trainFrameIndex) in enumerate(trainloader, 0):
## Getting the input and the target from the training set
frame0, frameT, frame1 = trainData
I0 = frame0.to(device)
I1 = frame1.to(device)
IFrame = frameT.to(device)
optimizer.zero_grad()
# Calculate flow between reference frames I0 and I1
flowOut = cnn1(torch.cat((I0, I1), dim=1))
# Extracting flows between I0 and I1 - F_0_1 and F_1_0
F_0_1 = flowOut[:, :2, :, :]
F_1_0 = flowOut[:, 2:, :, :]
fCoeff = model.getFlowCoeff(trainFrameIndex, device)
# Calculate intermediate flows
F_t_0 = fCoeff[0] * F_0_1 + fCoeff[1] * F_1_0
F_t_1 = fCoeff[2] * F_0_1 + fCoeff[3] * F_1_0
# Get intermediate frames from the intermediate flows
g_I0_F_t_0 = trainFlowBackWarp(I0, F_t_0)
g_I1_F_t_1 = trainFlowBackWarp(I1, F_t_1)
# Calculate optical flow residuals and visibility maps
intrpOut = cnn2(torch.cat((I0, I1, F_0_1, F_1_0, F_t_1, F_t_0, g_I1_F_t_1, g_I0_F_t_0), dim=1))
# Extract optical flow residuals and visibility maps
F_t_0_f = intrpOut[:, :2, :, :] + F_t_0
F_t_1_f = intrpOut[:, 2:4, :, :] + F_t_1
V_t_0 = torch.nn.functional.sigmoid(intrpOut[:, 4:5, :, :])
V_t_1 = 1 - V_t_0
# Get intermediate frames from the intermediate flows
g_I0_F_t_0_f = trainFlowBackWarp(I0, F_t_0_f)
g_I1_F_t_1_f = trainFlowBackWarp(I1, F_t_1_f)
wCoeff = model.getWarpCoeff(trainFrameIndex, device)
# Calculate final intermediate frame
Ft_p = (wCoeff[0] * V_t_0 * g_I0_F_t_0_f + wCoeff[1] * V_t_1 * g_I1_F_t_1_f) / (
wCoeff[0] * V_t_0 + wCoeff[1] * V_t_1)
interpolated_keypoints = []
intermediate_keypoints = []
for batchNum in range(2):
try:
TP(IFrame[batchNum].cpu().detach()).save("intermediate{}.jpg".format(batchNum))
datum = op.Datum()
image = cv.imread("./intermediate{}.jpg".format(batchNum))
datum.cvInputData = image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
if datum.poseKeypoints is None:
keypoints = np.zeros([25, 3], dtype=float)
else:
keypoints = datum.poseKeypoints[0]
intermediate_keypoints.append([keypoints])
except:
pass
for batchNum in range(2):
try:
TP(Ft_p[batchNum].cpu().detach()).save("interpolated{}.jpg".format(batchNum))
datum = op.Datum()
image = cv.imread("./interpolated{}.jpg".format(batchNum))
datum.cvInputData = image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
if datum.poseKeypoints is None:
keypoints = np.zeros([25, 3], dtype=float)
else:
keypoints = datum.poseKeypoints[0]
interpolated_keypoints.append([keypoints])
except:
pass
intermediate_keypoints = torch.tensor(intermediate_keypoints).float().to(device)
interpolated_keypoints = torch.tensor(interpolated_keypoints).float().to(device)
# Loss
recnLoss = L1_lossFn(Ft_p, IFrame)
prcpLoss = MSE_LossFn(vgg16_conv(Ft_p), vgg16_conv(IFrame))
warpLoss = L1_lossFn(g_I0_F_t_0, IFrame) + L1_lossFn(g_I1_F_t_1, IFrame) + L1_lossFn(
trainFlowBackWarp(I0, F_1_0), I1) + L1_lossFn(trainFlowBackWarp(I1, F_0_1), I0)
loss_smooth_1_0 = torch.mean(torch.abs(F_1_0[:, :, :, :-1] - F_1_0[:, :, :, 1:])) + torch.mean(
torch.abs(F_1_0[:, :, :-1, :] - F_1_0[:, :, 1:, :]))
loss_smooth_0_1 = torch.mean(torch.abs(F_0_1[:, :, :, :-1] - F_0_1[:, :, :, 1:])) + torch.mean(
torch.abs(F_0_1[:, :, :-1, :] - F_0_1[:, :, 1:, :]))
loss_smooth = loss_smooth_1_0 + loss_smooth_0_1
keypointsLoss = MSE_LossFn(interpolated_keypoints, intermediate_keypoints)
# code.interact(local=dict(globals(), **locals()))
# Total Loss - Coefficients 204 and 102 are used instead of 0.8 and 0.4
# since the loss in paper is calculated for input pixels in range 0-255
# and the input to our network is in range 0-1
loss = 163 * recnLoss + 82 * warpLoss + 0.004 * prcpLoss + 0.8 * loss_smooth + 0.001 * keypointsLoss
print("trainIndex: {}, loss: {}".format(trainIndex, loss))
# Backpropagate
loss.backward()
optimizer.step()
iLoss += loss.item()
# Validation and progress every 50 iterations
if ((trainIndex % 50) == 49):
end = time.time()
psnr, vLoss, valImg1, valImg2 = validate(validationloader)
valPSNR[epoch].append(psnr)
valLoss[epoch].append(vLoss)
# Tensorboard
itr = trainIndex + epoch * (len(trainloader))
writer.add_scalars('Loss', {'trainLoss': iLoss / 50,
'validationLoss': vLoss}, itr)
writer.add_scalar('PSNR', psnr, itr)
writer.add_image('Validation1', valImg1, itr)
writer.add_image('Validation2', valImg2, itr)
#####
endVal = time.time()
print(
" Loss: %0.6f Iterations: %4d/%4d TrainExecTime: %0.1f ValLoss:%0.6f ValPSNR: %0.4f ValEvalTime: %0.2f LearningRate: %f" % (
iLoss / 50, trainIndex, len(trainloader), end - start, vLoss, psnr, endVal - end,
get_lr(optimizer)))
cLoss[epoch].append(iLoss / 50)
iLoss = 0
start = time.time()
# Create checkpoint after every 3 epochs
if (not os.path.isdir('./checkpoints')):
os.mkdir('./checkpoints')
if ((epoch % 3) == 2):
dict1 = {
'Detail': "End to end Super SloMo.",
'epoch': epoch,
'timestamp': datetime.datetime.now(),
'trainBatchSz': 2,
'validationBatchSz': 2,
'learningRate': get_lr(optimizer),
'loss': cLoss,
'valLoss': valLoss,
'valPSNR': valPSNR,
'state_dictFC': cnn1.state_dict(),
'state_dictAT': cnn2.state_dict(),
}
torch.save(dict1, "./checkpoints/Interpolation" + str(checkpoint_counter) + ".ckpt")
checkpoint_counter += 1
def upload_file():
# result를 디비에 저장하기 위해 jwt 로 본인확인
current_user = get_jwt_identity()
print(current_user, '님이 분석을 요청하였습니다')
# 현재 시간
# 몽고 디비에 data 업데이트 할 때 사용
now = datetime.now()
# 이미지를 저장할 이름을 현재 시간으로 설정 XXXX.XXXX
image_save_name = datetime.timestamp(now)
file = request.files['video']
video_path = os.path.join(VIDEO_SAVE_PATH, f"{str(image_save_name)}.mp4")
# print('현재경로는: ', os.getcwd())
# print('비디오 파일 저장 경로는: ', video_path)
file.save(video_path)
print('비디오 저장')
"""
-----------------------
골프 db 에서 모델을 가져온다
-----------------------
"""
print('golfdb 시작')
ds = SampleVideo(video_path,
transform=transforms.Compose([
ToTensor(),
Normalize([0.485, 0.456, 0.406],
[0.229, 0.224, 0.225])
]))
dl = DataLoader(ds, batch_size=1, shuffle=False, drop_last=False)
model = EventDetector(pretrain=True,
width_mult=1.,
lstm_layers=1,
lstm_hidden=256,
bidirectional=True,
dropout=False)
save_dict = torch.load('models/golfdb/weight/swingnet_1800.pth.tar')
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)
model.load_state_dict(save_dict['model_state_dict'])
model.to(device)
model.eval()
print("Loaded model weights")
seq_length = 64
print('Testing...')
for sample in dl:
images = sample['images']
# full samples do not fit into GPU memory so evaluate sample in 'seq_length' batches
batch = 0
while batch * seq_length < images.shape[1]:
if (batch + 1) * seq_length > images.shape[1]:
image_batch = images[:, batch * seq_length:, :, :, :]
else:
image_batch = images[:, batch * seq_length:(batch + 1) *
seq_length, :, :, :]
logits = model(image_batch.cuda())
if batch == 0:
probs = F.softmax(logits.data, dim=1).cpu().numpy()
else:
probs = np.append(probs,
F.softmax(logits.data, dim=1).cpu().numpy(),
0)
batch += 1
events = np.argmax(probs, axis=0)[:-1]
print('Predicted event frames: {}'.format(events))
"""
cv2 비디오 캡쳐 오픈
"""
cap = cv2.VideoCapture(video_path)
confidence = []
for i, e in enumerate(events):
confidence.append(probs[e, i])
print('Condifence: {}'.format([np.round(c, 3) for c in confidence]))
"""
openpose 객체 오픈
및 파라미터 설정
"""
params = dict()
params["model_folder"] = "C:\\openpose\\models\\"
params["number_people_max"] = 1
params["net_resolution"] = "-1x240"
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
datum = op.Datum()
# 각 프레임의 키포인트를 dict 로 모음
key_data = dict()
for i, e in enumerate(events):
cap.set(cv2.CAP_PROP_POS_FRAMES, e)
_, img = cap.read()
# cv2.putText(img, '{:.3f}'.format(confidence[i]), (20, 20), cv2.FONT_HERSHEY_DUPLEX, 0.75, (0, 0, 255))
# golfdb 가 뽑아낸 이미지를 op 객체에
datum.cvInputData = img
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# yolo 로 뽑은 club head 데이터 추가
club_head_points, confiedences = yolo(frame=img,
size=416,
score_threshold=0.3,
nms_threshold=0.3)
x = club_head_points[0]
y = club_head_points[1]
w = club_head_points[2]
h = club_head_points[3]
club_head_list = np.array([x, y, confiedences])
_kp = datum.poseKeypoints[0]
key_data[i] = np.append(_kp, [club_head_list], axis=0)
# 오픈포즈 outputData에 클럽 헤드 좌표도 추가
output_image = cv2.rectangle(datum.cvOutputData, (x, y),
(x + w, y + h), (0, 0, 255), 2)
cv2.imwrite(f'{IMAGE_SAVE_PATH}{image_save_name}_{str(i)}.png',
output_image)
cap.release()
# 뽑아낸 키포인트를 분석
swing_anal = Anal.Anal(key_data, face_on=True)
result = swing_anal.check_all()
# result에 이미지가 저장되는 경로 넣어줌
result["filePath"] = image_save_name
# 결과 데이터에 업로드 날짜 추가
upload_date = now.strftime('%Y-%m-%d')
result["date"] = upload_date
# col.update_one(
# { "email": current_user},
# { "$push": {
# f"swingData": json.dumps(result, ensure_ascii=False)
# }
# }
# )
col.update_one(
{"email": current_user},
{"$push": {
"swingData": json.dumps(result, ensure_ascii=False)
}})
# print(json.dumps(result))
print(f'{current_user} : {upload_date}, 스윙 분석 업데이트')
# 분석 횟 수가 일정 횟수 이상이면 뱃지 발급
current_user_doc = col.find_one({"email": current_user})
counts = len(current_user_doc["swingData"])
print('분석 횟수는: ', counts)
if counts == 1:
col.update_one({'email': current_user},
{'$push': {
'badges': 'analysis_1'
}})
if counts == 10:
col.update_one({"email": current_user},
{"$push": {
"badges": "analysis_10"
}})
if counts == 30:
col.update_one({"email": current_user},
{"$push": {
"badges": "analysis_50"
}})
if counts == 50 and 'analysis_50' not in current_user_doc["badges"]:
col.update_one({"email": current_user},
{"$push": {
"badges": "analysis_50"
}})
# return json.dumps(result, cls=MyEncoder)
return result
def start(self):
# # load openpose python api
# if self.arg.openpose is not None:
# sys.path.append('{}/python'.format(self.arg.openpose))
# sys.path.append('{}/build/python'.format(self.arg.openpose))
# sys.path.append('{}/build/python/openpose'.format(self.arg.openpose))
# sys.path.append('{}/build/example/openpose'.format(self.arg.openpose))
# sys.path.append('{}/build/python/openpose/Release'.format(self.arg.openpose))
# sys.path.append('{}/build/x64/Release'.format(self.arg.openpose))
# sys.path.append('{}/build/bin'.format(self.arg.openpose))
# try:
# #from openpose import pyopenpose as op
# import pyopenpose as op
# except:
# print('Can not find Openpose Python API.')
# return
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
sys.path.append(dir_path +
'/../../openpose/build/python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + dir_path + '/../../openpose/build/x64/Release;' + dir_path + '/../../openpose/build/bin;'
import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
video_name = self.arg.video.split('/')[-1].split('.')[0]
label_name_path = './resource/kinetics_skeleton/label_name.txt'
with open(label_name_path) as f:
label_name = f.readlines()
label_name = [line.rstrip() for line in label_name]
self.label_name = label_name
# initiate
opWrapper = op.WrapperPython()
params = dict(model_folder='./models', model_pose='COCO')
opWrapper.configure(params)
opWrapper.start()
self.model.eval()
pose_tracker = naive_pose_tracker()
if self.arg.video == 'camera_source':
video_capture = cv2.VideoCapture(0)
else:
video_capture = cv2.VideoCapture(self.arg.video)
# start recognition
start_time = time.time()
frame_index = 0
while (True):
tic = time.time()
# get image
ret, orig_image = video_capture.read()
if orig_image is None:
break
source_H, source_W, _ = orig_image.shape
orig_image = cv2.resize(orig_image,
(256 * source_W // source_H, 256))
H, W, _ = orig_image.shape
# pose estimation
datum = op.Datum()
datum.cvInputData = orig_image
# opWrapper.emplaceAndPop([datum])
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
multi_pose = datum.poseKeypoints # (num_person, num_joint, 3) 关节点数据
# if (multi_pose is None) | (len(multi_pose.shape) != 3): # 是否三维数组
if multi_pose is None: # 是否三维数组
continue
# normalization
multi_pose[:, :, 0] = multi_pose[:, :, 0] / W
multi_pose[:, :, 1] = multi_pose[:, :, 1] / H
multi_pose[:, :, 0:2] = multi_pose[:, :, 0:2] - 0.5
multi_pose[:, :, 0][multi_pose[:, :, 2] == 0] = 0
multi_pose[:, :, 1][multi_pose[:, :, 2] == 0] = 0
# pose tracking
if self.arg.video == 'camera_source':
frame_index = int(
(time.time() - start_time) * self.arg.model_fps)
else:
frame_index += 1
pose_tracker.update(multi_pose, frame_index)
data_numpy = pose_tracker.get_skeleton_sequence()
data = torch.from_numpy(data_numpy)
data = data.unsqueeze(0)
data = data.float().to(
self.dev).detach() # (1, channel, frame, joint, person)
# model predict
voting_label_name, video_label_name, output, intensity = self.predict(
data)
# visualization
app_fps = 1 / (time.time() - tic)
image = self.render(data_numpy, voting_label_name,
video_label_name, intensity, orig_image,
app_fps)
cv2.imshow("ST-GCN", image)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
def detect_skeleton(file_name):
try:
# Import Openpose (Windows/Ubuntu/OSX)
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../openpose/build/python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + '../../openpose/build/x64/Release;' + '../../openpose/build/bin'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../openpose/build/python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
parser.add_argument("--video_path",
default=f"../media/{file_name}.mp4",
help="Read input video (avi, mp4).")
parser.add_argument("--no_display",
default=False,
help="Enable to disable the visual display.")
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "../../openpose/models/"
params["disable_multi_thread"] = "false"
numberGPUs = 1
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1:
next_item = args[1][i + 1]
else:
next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = next_item
# Construct it from system arguments
# op.init_argv(args[1])
# oppython = op.OpenposePython()
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Video
datum = op.Datum()
cap = cv2.VideoCapture(args[0].video_path)
set_frame_size(int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
# 관절을 입힌 동영상 생성
# fourcc = cv2.VideoWriter_fourcc('D', 'I', 'V', 'X')
# out = cv2.VideoWriter(f'../output/output/video/{file_name}.avi', fourcc, 20.0, get_frame_size())
frame_data = []
frame_id = -1
while cap.isOpened():
frame_id += 1
grabbed, frame = cap.read()
if frame is None or not grabbed:
print("Finish reading video frames...")
break
datum.cvInputData = frame
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# print("Body keypoints: \n" + str(datum.poseKeypoints))
# 동영상 저장
#out.write(datum.cvOutputData)
# cv2.imshow("OpenPose 1.7.0 - Tutorial Python API", datum.cvOutputData)
# cv2.waitKey(1)
# save the keypoint as a list
frame_data.append(make_json(datum, frame_id))
else:
print('cannot open the file')
# show list as json
with open(f'../output/json/output{file_name}.json',
'w',
encoding="utf-8") as make_file:
json.dump(frame_data, make_file, ensure_ascii=False, indent="\t")
cap.release()
# out.release()
# cv2.destroyAllWindows()
except Exception as e:
print(e)
sys.exit(-1)
return frame_data
def video_start():
global SSF_Wait, SLL_Wait, Pose, n, db, canvas, video_flag, cap, sx, sy, cursor, fill_line, canvas3, V
'''配置模型'''
params = dict() # 创建一个字典
params["model_folder"] = model # 修改路径
params["model_pose"] = "BODY_25" # 选择pose模型
params["number_people_max"] = 2 # 最大检测人体数
params["display"] = 2 # 2D模式
opWrapper = op.WrapperPython()
opWrapper.configure(params) # 导入上述参数
opWrapper.start()
'''配置模型'''
cap = cv2.VideoCapture(0) # 打开摄像头
Initial = 0
video_flag = True
week = str(strftime("%A", localtime()))
while video_flag: # 判断是否开启camera
SSF_Wait = 5
SLL_Wait = 2 # 加快video速度
cap.open(0)
ret, Frame = cap.read()
H = round(0.63 * sy)
W = round(0.906 * sy)
Frame = cv2.resize(Frame, (W, H)) # 图像尺寸变换
imageToProcess = cv2.cvtColor(Frame, cv2.COLOR_RGB2BGR) # RGB转换为BGR
BGR = imageToProcess
'''计算关节点'''
datum = op.Datum()
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# print("Body keypoints: \n" + str(datum.poseKeypoints))
Pose = datum.poseKeypoints # 关键点存在的地方,若无人像，输出None
# BGR = datum.cvOutputData # imageToProcess 已识别后的图像
'''计算关节点'''
'''初值'''
if Initial == 0: # 赋初值
'''SSF'''
n = 0 # 默认user=0
j = 0
j1 = 0
j2 = 0
j3 = 0 # 4种状态
SSF = 0 # 计数
SSF_Angle = 270 # 实际角度
SSF_Max = 270 # 抬手最大值
SSF_Time = 0 # 测量间隔
SSF_Num = 3 # 间隔数
SSF_Hold = 0 # hold时间
SSF_Flag = 0 # 计时阀门关闭
SSF_start = 0 # 开始时间
SSF_Deviation = 0 # 0.18s角度差值
SSF_Sum = 0 # 角度差值和
hip_angle = 90 # hip角度
helper = 'None'
string2 = 'None'
comps = 0 # 补偿角
SSF_Row = 1 # 第1行开始记录
'''SLL'''
i = 0
i1 = 0
i2 = 0
i3 = 0
SLL = 0
SLL_Angle = 120 # 默认120为开始lift leg
SLL_Max = 90
SLL_Time = 0 # 测量间隔
SLL_Num = 3 # 间隔数
SLL_Hold = 0 # hold时间
SLL_Flag = 0 # 计时阀门关闭
SLL_start = 0 # 开始时间
SLL_Deviation = 0 # 0.18s角度差值
SLL_Sum = 0 # 角度差值和
SLL_Row = 1 # 第1行开始记录
SLL_string2 = 'None'
# 实际角度
V = 1
Initial = 1 # 赋值完毕
'''初值'''
'''模型测量'''
if str(Pose) != 'None':
'''Helper'''
if Pose.shape[0] == 1:
n = 0 # Only one person
helper = 'Yes!'
else:
if Pose[1][8][2] != 0 and Pose[1][9][2] != 0 and Pose[1][12][2] != 0:
if Pose[0][8][2] != 0 and Pose[0][9][2] != 0 and Pose[0][12][2] != 0:
if Pose[0][8][1] > Pose[1][8][1] and Pose[0][9][1] > Pose[1][9][1] and Pose[0][12][1] > \
Pose[1][12][1]:
n = 0 # 0先生是坐着的
else:
n = 1
else:
n = 1
else:
n = 0
helper = 'No'
'''Helper''' # n = user
'''Compensation'''
# neck_angle = round(Angle(5, 1, 8), 0)
comps = np.arctan(abs((Pose[n][1][0] - Pose[n][8][0])) / (Pose[n][8][1] - Pose[n][1][1])) * 180 / np.pi
comps = round(comps, 0)
if comps < 10:
string2 = 'Straight!'
comps = 0
else:
if Pose[n][1][0] > Pose[n][8][0]:
string2 = 'Left Comps! ' + str(comps)
else:
string2 = 'Right Comps! ' + str(abs(comps))
if SSF_Max > Angle(1, 2, 3) and j == 1 and j1 == 1 and j2 == 1 and j3 == 1:
SSF_Max = round(Angle(1, 2, 3), 0) # 抬手时找角度最小
# real_angle > Angle(1, 2, 3) - (180 - hip_angle - neck_angle)
'''Compensation'''
'''SSF'''
if Pose[n][4][2] != 0 and Pose[n][2][2] != 0 and Pose[n][1][2] != 0 and Pose[n][3][2] != 0:
'''Num'''
SSF_Deviation = abs(SSF_Angle - Angle(1, 2, 3))
SSF_Angle = Angle(1, 2, 3)
if Pose[n][4][1] < Pose[n][2][1]: # hands up
j3 = 1
else: # hands down
j3 = 0
if j == 0 and j1 == 0 and j2 == 1 and j3 == 1: # 0011抬手
SSF = SSF + 1
j = j1
j1 = j2
j2 = j3
'''Num'''
'''Hold time'''
if j == 1 and j1 == 1 and j2 == 1 and j3 == 1: # 1111 hands up
SSF_Time = SSF_Time + 1 # 单位0.18s
if SSF_Time % SSF_Num == 0: # 开始判定
if SSF_Sum / SSF_Num < 6: # if hold？
if SSF_Flag == 0:
SSF_start = time()
SSF_Flag = 1
else:
SSF_Flag = 0
SSF_Sum = 0
else: # 角度均值
SSF_Sum = SSF_Deviation + SSF_Sum
else:
SSF_Flag = 0
if SSF_Flag == 1 and round(time() - SSF_start, 2) > 1: # 找到最后的hold值
SSF_Hold = round(time() - SSF_start, 2)
'''Hold time'''
'''Fill in database'''
if j == 1 and j1 == 1 and j2 == 0 and j3 == 0: # 1100放手
sql = "INSERT INTO " + week + "(SSF_Times, SSF_Angle, SSF_Hold, SSF_Comps ,SSF_DIY) VALUES (%s, %s, " \
"%s, %s, %s); "
cursor.execute(sql, (SSF_Row, 270 - SSF_Max, SSF_Hold, comps, helper))
db.commit()
SSF_Row = SSF_Row + 1
SSF_Max = 270
SSF_Hold = 0
'''Fill in database'''
else:
BGR = cv2.putText(BGR, 'Move left', (5, H - 90), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
'''SSF'''
###################################################################################################
'''Sitting Leg Lift'''
if Pose[n][9][2] != 0 and Pose[n][10][2] != 0 and Pose[n][11][2] != 0:
'''COMPS'''
SLL_comps = np.arctan(
abs((Pose[n][9][1] - Pose[n][10][1]) / (Pose[n][10][0] - Pose[n][9][0]))) * 180 / np.pi
SLL_comps = round(SLL_comps, 0)
if SLL_comps < 15:
SLL_string2 = 'Horizontal'
SLL_comps = 0
else:
SLL_string2 = 'Comps! ' + str(abs(SLL_comps))
'''COMPS'''
'''Num'''
SLL_Deviation = SLL_Angle - Angle(9, 10, 11)
SLL_Angle = Angle(9, 10, 11)
if SLL_Max < SLL_Angle:
SLL_Max = round(SLL_Angle, 0)
if SLL_Angle > 120: # leg up
i3 = 1
else: # leg down
i3 = 0
if i == 0 and i1 == 0 and i2 == 1 and i3 == 1: # 0011抬脚
SLL = SLL + 1
i = i1
i1 = i2
i2 = i3
'''Num'''
'''Hold time'''
if i == 1 and i1 == 1 and i2 == 1 and i3 == 1: # 1111 hands up
SLL_Time = SLL_Time + 1 # 单位0.18s
if SLL_Time % SLL_Num == 0: # 开始判定
if SLL_Sum / SSF_Num < 6: # if hold？
if SLL_Flag == 0:
SLL_start = time()
SLL_Flag = 1
else:
SLL_Flag = 0
SLL_Sum = 0
else: # 角度偏差均值
SLL_Sum = SLL_Deviation + SLL_Sum
else:
SLL_Flag = 0
if SLL_Flag == 1 and round(time() - SLL_start, 2) > 1: # 找到最后的hold值
SLL_Hold = round(time() - SLL_start, 2)
'''Hold time'''
'''Fill in database'''
if i == 1 and i1 == 1 and i2 == 0 and i3 == 0: # 1100放腿
sql = "INSERT INTO " + week + "(SLL_Times, SLL_Angle, SLL_Hold, SLL_Comps ,SLL_DIY) VALUES (%s, %s, " \
"%s, %s, %s); "
cursor = db.cursor()
cursor.execute(sql, (SLL_Row, SLL_Max, SLL_Hold, SLL_comps, helper))
db.commit()
SLL_Row = SLL_Row + 1
SLL_Max = 90
SLL_Hold = 0
'''Fill in database'''
else:
BGR = cv2.putText(BGR, 'Move left', (5, H - 90), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
'''SLL'''
else:
BGR = cv2.putText(BGR, 'Move into camera', (5, H - 60), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
'''mirror_feedback'''
string0 = 'SSF: ' + str(SSF) + ' Hold:' + str(SSF_Hold) + 's'
BGR = cv2.putText(BGR, string0, (5, 30), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
BGR = cv2.putText(BGR, 'Angle: ' + str(270 - SSF_Max), (5, 70), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
BGR = cv2.putText(BGR, string2, (5, 110), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
BGR = cv2.putText(BGR, helper, (5, 150), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (255, 0, 0), 2)
############################################################################################
SLL_string0 = 'SLL: ' + str(SLL) + ' Hold:' + str(SLL_Hold) + 's'
BGR = cv2.putText(BGR, SLL_string0, (W - 350, 30), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
BGR = cv2.putText(BGR, 'Angle: ' + str(SLL_Max), (W - 350, 70), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
BGR = cv2.putText(BGR, SLL_string2, (W - 350, 110), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
BGR = cv2.putText(BGR, helper, (W - 350, 150), cv2.FONT_HERSHEY_SIMPLEX, 1.2, (0, 0, 255), 2)
'''控件显示'''
# mirror feedback
RGB = Image.fromarray(BGR)
RGB = ImageTk.PhotoImage(RGB)
canvas.create_image(0, 0, anchor='nw', image=RGB) # 刷新图片，nw表示从西北角开始排列图片
canvas.bg = RGB
text2.set('Camera Open!')
# times
print(SSF, V)
if int(SSF) < int(V):
text3.set(str(SSF) + ' / ' + str(V))
else:
text3.set('Finish!')
# progress bar
if j == 1 and j1 == 1 and j2 == 1 and j3 == 1: # hands up
k = 0.4 * sx / 90 # 总长度0.4*sx
n = k * (180 - SSF_Max)
else:
n = 0
canvas3.coords(fill_line, (0, 0, n, 60))
window.update()
cv2.waitKey(1)
def pose_estimation(self):
# load openpose python api
if self.arg.openpose is not None:
# sys.path.append('{}/python'.format(self.arg.openpose))
# sys.path.append('{}/build/python'.format(self.arg.openpose))
# print(self.arg.openpose + '/build/python/openpose/Release')
sys.path.append(self.arg.openpose +
'/build/python/openpose/Release')
# sys.path.append(self.arg.openpose + '\\build\\x64\\Release')
# sys.path.append(self.arg.openpose + '\\build\\bin')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + self.arg.openpose + '/build/x64/Release;' + self.arg.openpose + '/build/bin;'
print(sys.path)
try:
# from openpose import pyopenpose as op
import pyopenpose as op
except:
print('Can not find Openpose Python API.')
return
video_name = self.arg.video.split('/')[-1].split('.')[0]
# initiate
opWrapper = op.WrapperPython()
params = dict(model_folder='./models', model_pose='COCO')
opWrapper.configure(params)
opWrapper.start()
self.model.eval()
video_capture = cv2.VideoCapture(self.arg.video)
video_length = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
pose_tracker = naive_pose_tracker(data_frame=video_length)
# pose estimation
start_time = time.time()
frame_index = 0
video = list()
pose_list = []
pbar = tqdm.tqdm(total=video_length)
while (True):
# get image
ret, orig_image = video_capture.read()
if orig_image is None:
break
source_H, source_W, _ = orig_image.shape
orig_image = cv2.resize(orig_image,
(self.arg.shortedge, self.arg.shortedge *
source_H // source_W))
H, W, _ = orig_image.shape
video.append(orig_image)
# pose estimation
datum = op.Datum()
datum.cvInputData = orig_image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
multi_pose = datum.poseKeypoints # (num_person, num_joint, 3)
if len(multi_pose.shape) != 3:
continue
my_pose = multi_pose.copy()
my_pose = self.__get_prior_person(my_pose)
pose_list.append(my_pose)
# normalization
multi_pose[:, :, 0] = multi_pose[:, :, 0] / W
multi_pose[:, :, 1] = multi_pose[:, :, 1] / H
multi_pose[:, :, 0:2] = multi_pose[:, :, 0:2] - 0.5
multi_pose[:, :, 0][multi_pose[:, :, 2] == 0] = 0
multi_pose[:, :, 1][multi_pose[:, :, 2] == 0] = 0
# pose tracking
pose_tracker.update(multi_pose, frame_index)
frame_index += 1
pbar.update(1)
# print('Pose estimation ({}/{}).'.format(frame_index, video_length))
data_numpy = pose_tracker.get_skeleton_sequence()
pbar.close()
# np.save('stgcn_data_numpy_original.npy', data_numpy)
return video, data_numpy, pose_list
def toCsv(file_path):
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../python/openpose/Release')
os.environ['PATH'] = os.environ['PATH'] + ';' + dir_path + \
'/../../x64/Release;' + dir_path + '/../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('/home/aleix/openpose/build/python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
parser.add_argument(
"--video_path",
default=file_path,
help=
"Process an image. Read all standard formats (jpg, png, bmp, etc.).")
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "/home/aleix/openpose/models/"
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1:
next_item = args[1][i + 1]
else:
next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = next_item
try:
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
datum = op.Datum()
imageToProcess = cv2.VideoCapture(args[0].video_path)
arrayNoNorm = np.array(0)
totalFrames = int(imageToProcess.get(cv2.CAP_PROP_FRAME_COUNT))
target = cutFps(totalFrames)
counter = 0
framesRet = 0
# 1. Guardem totes les posicions del vídeo en una array.
while (imageToProcess.isOpened()):
if counter == target:
ret, frame = imageToProcess.read()
if ret == True:
datum.cvInputData = frame
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# Display Image
kp_array = np.array(datum.poseKeypoints)
# Treiem els outliers
nou_array = np.delete(kp_array[0], [3, 4, 15, 17, 22, 23],
0)
arrayNoNorm = np.append(arrayNoNorm, nou_array)
# print("Body keypoints: \n" + str(datum.poseKeypoints))
# cv2.imshow("OpenPose 1.7.0 - Tutorial Python API",
# datum.cvOutputData)
cv2.waitKey(1)
counter = 0
framesRet += 1
else:
break
else:
imageToProcess.grab()
counter += 1
label = file_path.split('/')[7]
label = label.split('_')[0]
label = label.split('.')[0]
print(label)
#label = label.join([c for c in label if c.isalpha()])
pos = 0
# print("Frames sencers: ",totalFrames)
# print("Frames retallats: ", framesRet)
for i in range(framesRet):
# 2. Normalitzem dades frame per frame (cada 75 posicions ja que te x, y i accuracy)
posini = pos
pos += 57
arrayFrame = arrayNoNorm[posini:pos]
# print(len(arrayFrame))
# if len(arrayFrame) == 1:
# print(arrayFrame)
arrayFrame = normValues(arrayFrame, label)
# 3. Les guardem en un csv
with open(
'/home/aleix/Escriptori/coses_tfg/tfg-correct-exercise/api/files/keypoints.csv',
'a') as file:
writer = csv.writer(file)
writer.writerow(arrayFrame)
# Fem padding fins a 75, ja que és el màxim de frames.
padding = [label]
for i in range(38):
padding.append(-1000.0)
while (framesRet != 75):
with open(
'/home/aleix/Escriptori/coses_tfg/tfg-correct-exercise/api/files/keypoints.csv',
'a') as file:
writer = csv.writer(file)
writer.writerow(padding)
framesRet += 1
except Exception as e:
print(e)
sys.exit(-1)
def toCsv(file_path):
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../python/openpose/Release')
os.environ['PATH'] = os.environ['PATH'] + ';' + dir_path + \
'/../../x64/Release;' + dir_path + '/../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
parser.add_argument(
"--image_path",
default=file_path,
help=
"Process an image. Read all standard formats (jpg, png, bmp, etc.).")
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "../../../models/"
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1:
next_item = args[1][i + 1]
else:
next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = next_item
try:
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
datum = op.Datum()
imageToProcess = cv2.imread(args[0].image_path)
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# Display Image
kp_array = np.array(datum.poseKeypoints)
# 1. Guardem totes les posicions del vídeo en una array.
arrayFrame = normValues(kp_array)
# print(arrayFrame)
with open(
'/home/aleix/Escriptori/coses_tfg/tfg-correct-exercise/dataset/train_dataset.csv',
'a') as file:
writer = csv.writer(file)
writer.writerow(arrayFrame)
except Exception as e:
print(e)
sys.exit(-1)
def main():
buildPath = os.path.dirname(os.path.realpath(__file__))
sys.path.append(buildPath + "/python/openpose/Release");
os.environ['PATH'] += ';' + (buildPath + "/x64/Release") + ';' + (buildPath + "/bin;")
import pyopenpose as op
params = dict()
params["model_folder"] = "../models/"
params["model_pose"] = "BODY_25"
params["net_resolution"] = "-1x224"
params["hand"] = True
params["hand_detector"] = 0
params["hand_net_resolution"] = "224x224"
openposeHandle = op.WrapperPython()
openposeHandle.configure(params)
openposeHandle.start()
datum = op.Datum()
cam = cv2.VideoCapture(0)
frameElapsedTime = 1
totalTimeElapsed = 0
sampleIndex = 0
OutputStr = ""
while True:
frameStart = time.time()
ret, frame = cam.read()
frame = cv2.flip(frame, 1)
datum.cvInputData = frame
openposeHandle.emplaceAndPop(op.VectorDatum([datum]))
if type(datum.poseKeypoints) != type(None):
if totalTimeElapsed > samplePeriod:
totalTimeElapsed -= samplePeriod
focusIndex = getFocusBodyIndex(datum.poseKeypoints, len(datum.poseKeypoints))
body = datum.poseKeypoints[focusIndex]
leftHand = datum.handKeypoints[0][focusIndex]
rightHand = datum.handKeypoints[1][focusIndex]
X=[body[0][0], body[0][1], body[1][0], body[1][1], body[2][0], body[2][1], body[3][0], body[3][1],body[4][0],
body[4][1], body[5][0], body[5][1], body[6][0], body[6][1], body[7][0], body[7][1],body[15][0],
body[15][1], body[16][0], body[16][1], body[17][0], body[17][1], body[18][0], body[18][1],leftHand[0][0], leftHand[0][1]]
Xdict={'a':X}
X= pd.DataFrame(Xdict)
#X=preprocessing.scale(X)
#print(X)
preX=detect.predict(X.T)
max_a = np.argmax(preX, axis=1)
if max_a==0:
OutputStr = "studying"
if max_a==1:
OutputStr = "using smartphone"
if max_a==2:
OutputStr = "sleeping"
datum.cvOutputData = cv2.putText(datum.cvOutputData, OutputStr, (2, 34), cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0))
datum.cvOutputData = cv2.putText(datum.cvOutputData,OutputStr, (0, 32), cv2.FONT_HERSHEY_DUPLEX, 1, (255, 255, 255))
#datum.cvOutputData = cv2.putText(datum.cvOutputData, "FPS: %.02f Target FPS: %.02f" % (1.0 / frameElapsedTime, frameRate), (2, 34), cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0))
#datum.cvOutputData = cv2.putText(datum.cvOutputData,"FPS: %.02f Target FPS: %.02f" % (1.0 / frameElapsedTime, frameRate), (0, 32), cv2.FONT_HERSHEY_DUPLEX, 1, (255, 255, 255))
cv2.imshow(windowName, datum.cvOutputData)
frameEnd = time.time()
remainTime = (1.0 / frameRate) - (frameEnd - frameStart)
if remainTime > 0:
key = cv2.waitKey(int(remainTime*1000) & 0xffffffff)
else:
key = cv2.waitKey(1)
frameEnd = time.time()
frameElapsedTime = frameEnd - frameStart
totalTimeElapsed += frameElapsedTime
if cv2.getWindowProperty(windowName, cv2.WND_PROP_ASPECT_RATIO) == -1: break
cam.release()
def image_preprocess(file_path):
try:
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + dir_path + '/../../x64/Release;' + dir_path + '/../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
parser.add_argument(
"--image_path",
default=file_path,
help=
"Process an image. Read all standard formats (jpg, png, bmp, etc.)."
)
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "../../../models/"
params["write_images"] = "./src/temp"
params["write_images_format"] = "png"
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1:
next_item = args[1][i + 1]
else:
next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = next_item
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
datum = op.Datum()
imageToProcess = cv2.imread(args[0].image_path)
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# Calculate Angles
data = {
'kneeHipAngle':
calculate_angle(datum.poseKeypoints[0][12],
datum.poseKeypoints[0][13],
datum.poseKeypoints[0][14]),
'hipChestAngle':
calculate_angle(datum.poseKeypoints[0][5],
datum.poseKeypoints[0][12],
datum.poseKeypoints[0][13]),
'chestArmAngle':
calculate_angle(datum.poseKeypoints[0][12],
datum.poseKeypoints[0][5],
datum.poseKeypoints[0][6]),
'armAngleDiff':
calculate_angle(datum.poseKeypoints[0][4],
datum.poseKeypoints[0][5],
datum.poseKeypoints[0][7]),
'kneeAngleDiff':
calculate_angle(datum.poseKeypoints[0][11],
datum.poseKeypoints[0][13],
datum.poseKeypoints[0][14])
}
return data
except Exception as e:
print(e)
sys.exit(-1)
for imageBaseId in range(0, len(imagePaths), numberGPUs):
# Create datums
images = []
# Read and push images into OpenPose wrapper
for gpuId in range(0, numberGPUs):
imageId = imageBaseId + gpuId
if imageId < len(imagePaths):
imagePath = imagePaths[imageBaseId + gpuId]
datum = op.Datum()
images.append(cv2.imread(imagePath))
datum.cvInputData = images[-1]
opWrapper.waitAndEmplace(op.VectorDatum([datum]))
# Retrieve processed results from OpenPose wrapper
for gpuId in range(0, numberGPUs):
imageId = imageBaseId + gpuId
if imageId < len(imagePaths):
datums = op.VectorDatum()
opWrapper.waitAndPop(datums)
datum = datums[0]
print("Body keypoints: \n" + str(datum.poseKeypoints))
if not args[0].no_display:
cv2.imshow("OpenPose 1.7.0 - Tutorial Python API",
def jsonify(file_path):
try:
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../python/openpose/Release')
os.environ['PATH'] = os.environ['PATH'] + ';' + dir_path + \
'/../../x64/Release;' + dir_path + '/../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append('../../python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
# sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
parser.add_argument(
"--video_path",
default=file_path,
help=
"Process an image. Read all standard formats (jpg, png, bmp, etc.)."
)
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "../../../models/"
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1:
next_item = args[1][i + 1]
else:
next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params:
params[key] = next_item
base = os.path.basename(file_path)
nom = os.path.splitext(base)[0]
result_dir = "/home/aleix/Escriptori/coses_tfg/tfg-correct-exercise/dataset/"
nom_fitxer = result_dir + nom + ".json"
kplist = [] # Llista de keypoints
kpdict = {} # Diccionari de keypoints
numLinia = 1 # Indica el número de la línia en que està al fitxer
vegada = 0 # Indica la a quin punt del cos està
nextFrame = False # Indica quan passa a un altre frame
totalFrames = 0
try:
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
datum = op.Datum()
imageToProcess = cv2.VideoCapture(args[0].video_path)
while (imageToProcess.isOpened()):
ret, frame = imageToProcess.read()
if ret == True:
datum.cvInputData = frame
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
totalFrames += 1
# Display Image
kpdict['body keypoint'] = np.array(
datum.poseKeypoints).tolist()
kplist.append(kpdict.copy())
# print("Body keypoints: \n" + str(datum.poseKeypoints))
cv2.imshow("OpenPose 1.7.0 - Tutorial Python API",
datum.cvOutputData)
cv2.waitKey(1)
else:
break
except Exception as e:
print(e)
sys.exit(-1)
kplist.append("Frames: " + str(totalFrames))
f = open(nom_fitxer, "w+")
json.dump(kplist, f, indent=4)
f.close()
with open(nom_fitxer) as fp:
linies = fp.read().splitlines()
with open(nom_fitxer, "w") as fp:
for linia in linies:
if nextFrame:
print(linia, file=fp)
if numLinia == 10:
nextFrame = False
numLinia = 5
vegada = 0
else:
numLinia += 1
elif numLinia % 5 == 0:
if vegada == 0:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Nas",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 1:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Clatell",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 2:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Espatlla dreta",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 3:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Colze dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 4:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Canell dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 5:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Espatlla esquerra",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 6:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Colze esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 7:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Canell esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 8:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Cadera centre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 9:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Cadera banda dreta",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 10:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Genoll dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 11:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Turmell dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 12:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Cadera banda esquerra",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 13:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Genoll esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 14:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Turmell esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 15:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Ull dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 16:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Ull esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 17:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Orella dreta",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 18:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Orella esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 19:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Dit gros esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 20:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Dit petit esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 21:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Taló esquerre",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 22:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Dit gros dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 23:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Dit petit dret",',
file=fp)
vegada += 1
numLinia += 1
elif vegada == 24:
print(linia + "\n" + "\t" + "\t" + "\t" + "\t" + "\t" +
'"Taló dret",',
file=fp)
numLinia = 1
nextFrame = True
else:
print(linia, file=fp)
numLinia += 1
fp.close()
except Exception as e:
print(e)
sys.exit(-1)
def pose_estimation(self): # 调用openpose识别出关节点
# load openpose python api
# if self.arg.openpose is not None:
# sys.path.append('{}/python'.format(self.arg.openpose))
# sys.path.append('{}/build/python'.format(self.arg.openpose))
# sys.path.append('{}/build/python/openpose'.format(self.arg.openpose))
# sys.path.append('{}/build/example/openpose'.format(self.arg.openpose))
# sys.path.append('{}/build/python/openpose/Release'.format(self.arg.openpose))
# sys.path.append('{}/build/x64/Release'.format(self.arg.openpose))
# sys.path.append('{}/build/bin'.format(self.arg.openpose))
# try:
# #from openpose import pyopenpose as op
# import pyopenpose as op
# except:
# print('Can not find Openpose Python API.')
# return
# dir_path = os.path.dirname(os.path.realpath(__file__))
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
# if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
# sys.path.append('D:/PycharmProject/openpose/build/python/openpose/Release')
# os.environ['PATH'] = os.environ['PATH'] + ';' + dir_path + 'D:/PycharmProject/openpose/build/x64/Release' + dir_path + 'D:/PycharmProject/openpose/build/bin'
sys.path.append(dir_path +
'/../../openpose/build/python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + dir_path + '/../../openpose/build/x64/Release;' + dir_path + '/../../openpose/build/bin;'
import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
video_name = self.arg.video.split('/')[-1].split('.')[0]
# initiate
opWrapper = op.WrapperPython()
params = dict(model_folder='./models', model_pose='COCO')
opWrapper.configure(params)
opWrapper.start()
self.model.eval()
video_capture = cv2.VideoCapture(self.arg.video)
# video_capture = cv2.VideoCapture(0) ＃ ０为打开摄像头
video_length = int(video_capture.get(cv2.CAP_PROP_FRAME_COUNT))
pose_tracker = naive_pose_tracker(data_frame=video_length)
# pose estimation
start_time = time.time()
frame_index = 0
video = list()
while (True):
# get image
ret, orig_image = video_capture.read()
if orig_image is None:
break
source_H, source_W, _ = orig_image.shape
orig_image = cv2.resize(orig_image,
(256 * source_W // source_H, 256))
H, W, _ = orig_image.shape
video.append(orig_image)
# pose estimation
datum = op.Datum()
datum.cvInputData = orig_image
# opWrapper.emplaceAndPop([datum])
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
multi_pose = datum.poseKeypoints # (num_person, num_joint, 3)
#if len(multi_pose.shape) != 3: # 是否三维数组
if (multi_pose is None): # 是否有数据，否则为None的时候会报错
continue
# normalization
multi_pose[:, :, 0] = multi_pose[:, :, 0] / W
multi_pose[:, :, 1] = multi_pose[:, :, 1] / H
multi_pose[:, :, 0:2] = multi_pose[:, :, 0:2] - 0.5
multi_pose[:, :, 0][multi_pose[:, :, 2] == 0] = 0
multi_pose[:, :, 1][multi_pose[:, :, 2] == 0] = 0
# pose tracking
pose_tracker.update(multi_pose, frame_index)
frame_index += 1
print('Pose estimation ({}/{}).'.format(frame_index, video_length))
data_numpy = pose_tracker.get_skeleton_sequence()
return video, data_numpy
def main():
try:
# Import Openpose (Windows/Ubuntu/OSX)
dir_path = os.path.dirname(os.path.realpath(__file__))
try:
# Windows Import
if platform == "win32":
# Change these variables to point to the correct folder (Release/x64 etc.)
sys.path.append(dir_path + '/../../python/openpose/Release')
os.environ['PATH'] = os.environ[
'PATH'] + ';' + dir_path + '/../../x64/Release;' + dir_path + '/../../bin;'
import pyopenpose as op
else:
# Change these variables to point to the correct folder (Release/x64 etc.)
# sys.path.append('/usr/local/python')
# If you run `make install` (default path is `/usr/local/python` for Ubuntu), you can also access the OpenPose/python module from there. This will install OpenPose and the python library at your desired installation path. Ensure that this is in your python path in order to use it.
sys.path.append('/usr/local/python')
from openpose import pyopenpose as op
except ImportError as e:
print(
'Error: OpenPose library could not be found. Did you enable `BUILD_PYTHON` in CMake and have this Python script in the right folder?'
)
raise e
# Flags
parser = argparse.ArgumentParser()
args = parser.parse_known_args()
# Custom Params (refer to include/openpose/flags.hpp for more parameters)
params = dict()
params["model_folder"] = "/home/medrobotics/openpose/models/"
params["net_resolution"] = "-1x352"
#params["face"] = True
#params["face_detector"] = 2
# Add others in path?
for i in range(0, len(args[1])):
curr_item = args[1][i]
if i != len(args[1]) - 1: next_item = args[1][i + 1]
else: next_item = "1"
if "--" in curr_item and "--" in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = "1"
elif "--" in curr_item and "--" not in next_item:
key = curr_item.replace('-', '')
if key not in params: params[key] = next_item
# Construct it from system arguments
# op.init_argv(args[1])
# oppython = op.OpenposePython()
# Starting OpenPose
opWrapper = op.WrapperPython()
opWrapper.configure(params)
opWrapper.start()
# Process Image
for i in range(10000):
datum = op.Datum()
string = '/media/medrobotics/STARS_B/2021_12_20_A/color/' + str(
i) + '.bin'
imageToProcess = np.fromfile(string, dtype=np.uint8)
imageToProcess = np.reshape(imageToProcess, (1080, 1920, 4))
imageToProcess = imageToProcess[:, :, :3].copy()
datum.cvInputData = imageToProcess
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
out = datum.cvOutputData
if datum.poseKeypoints is not None:
for j in range(datum.poseKeypoints.shape[0]):
mean = np.zeros(shape=(2, ))
variance = np.zeros(shape=(2, ))
count = 0
for k in [0, 14, 15, 16, 17]:
if datum.poseKeypoints[j][k][2] != 0.0:
count += 1
if count == 0:
break
for k in [0, 14, 15, 16, 17]:
if datum.poseKeypoints[j][k][2] != 0.0:
mean += datum.poseKeypoints[j][k][:2] / count
for k in [0, 14, 15, 16, 17]:
if datum.poseKeypoints[j][k][2] != 0.0:
variance += (datum.poseKeypoints[j][k][:2] -
mean) * (datum.poseKeypoints[j][k][:2]
- mean) / count
std_dev = np.sqrt(variance)
std_dev[0] = 10
std_dev[1] = 10
x1 = min(max(int(mean[0] - 3 * std_dev[0]), 0), 1920)
x2 = min(max(int(mean[0] + 3 * std_dev[0]), 0), 1920)
y1 = min(max(int(mean[1] - 3 * std_dev[1]), 0), 1080)
y2 = min(max(int(mean[1] + 3 * std_dev[1]), 0), 1080)
face = out[y1:y2, x1:x2]
#if face is not None and face.size != 0:
#face = cv2.GaussianBlur(face, (101, 101), 100)
#out[y1:y2, x1:x2] = face
cv2.imshow("OpenPose 1.7.0 - Tutorial Python API", out)
if cv2.waitKey(1) == ord('q'):
sys.exit(1)
except Exception as e:
print(e)
sys.exit(-1)
if lrdiff < btdiff:
len = lrdiff
y = y + (btdiff - lrdiff) / 2
else:
len = btdiff
x = x + (lrdiff - btdiff) / 2
faceRectangles = [
op.Rectangle(x, y, len, len),
]
# Create new datum
datum = op.Datum()
datum.cvInputData = imageToProcess
datum.faceRectangles = faceRectangles
# Process and display image
opWrapper.emplaceAndPop(op.VectorDatum([datum]))
# print("Face keypoints: \n" + str(datum.faceKeypoints))
# save openpose data
# rootForFacepos = root.replace('iphone_pictures', 'openpose')
# if not isdir(rootForFacepos):
# os.makedirs(rootForFacepos)
# fileForFacepos = file.replace('jpg', 'npy')
# fileSave = rootForFacepos + '/' + fileForFacepos
# np.save(fileSave, datum.faceKeypoints)
imS = cv2.resize(datum.cvOutputData, (540, 960))
cv2.imshow("OpenPose 1.7.0 - Tutorial Python API", imS)
cv2.waitKey(0)
def main():
buildPath = os.path.dirname(os.path.realpath(__file__))
t = datetime.datetime.now()
samplePath = buildPath + "/my_samples_%04d_%02d_%02d_%02d_%02d_%02d" % (
t.year, t.month, t.day, t.hour, t.minute, t.second)
sys.path.append(buildPath + "/python/openpose/Release")
os.environ['PATH'] += ';' + (buildPath +
"/x64/Release") + ';' + (buildPath + "/bin;")
import pyopenpose as op
params = dict()
params["model_folder"] = "../models/"
params["model_pose"] = "BODY_25"
params["net_resolution"] = "-1x160"
params["hand"] = True
params["hand_detector"] = 0
params["hand_net_resolution"] = "224x224"
openposeHandle = op.WrapperPython()
openposeHandle.configure(params)
openposeHandle.start()
os.makedirs(samplePath)
datum = op.Datum()
cam = cv2.VideoCapture(0)
with open(samplePath + "/sample_coords.csv", 'w', newline='') as csvFile:
csvWriter = csv.writer(csvFile)
csvWriter.writerow([
"focusing", "nose_x", "nose_y", "neck_x", "neck_y", "rShoulder_x",
"rShoulder_y", "rElbow_x", "rElbow_y", "rWrist_x", "rWrist_y",
"lShoulder_x", "lShoulder_y", "lElbow_x", "lElbow_y", "lWrist_x",
"lWrist_y", "rEye_x", "rEye_y", "lEye_x", "lEye_y", "rEar_x",
"rEar_y", "lEar_x", "lEar_y", "lHand_x", "lHand_y", "lThumb0_x",
"lThumb0_y", "lThumb1_x", "lThumb1_y", "lThumb2_x", "lThumb2_y",
"lThumb3_x", "lThumb3_y", "lIndexFinger0_x", "lIndexFinger0_y",
"lIndexFinger1_x", "lIndexFinger1_y", "lIndexFinger2_x",
"lIndexFinger2_y", "lIndexFinger3_x", "lIndexFinger3_y",
"lMiddleFinger0_x", "lMiddleFinger0_y", "lMiddleFinger1_x",
"lMiddleFinger1_y", "lMiddleFinger2_x", "lMiddleFinger2_y",
"lMiddleFinger3_x", "lMiddleFinger3_y", "lRingFinger0_x",
"lRingFinger0_y", "lRingFinger1_x", "lRingFinger1_y",
"lRingFinger2_x", "lRingFinger2_y", "lRingFinger3_x",
"lRingFinger3_y", "lLittleFinger0_x", "lLittleFinger0_y",
"lLittleFinger1_x", "lLittleFinger1_y", "lLittleFinger2_x",
"lLittleFinger2_y", "lLittleFinger3_x", "lLittleFinger3_y",
"rHand_x", "rHand_y", "rThumb0_x", "rThumb0_y", "rThumb1_x",
"rThumb1_y", "rThumb2_x", "rThumb2_y", "rThumb3_x", "rThumb3_y",
"rIndexFinger0_x", "rIndexFinger0_y", "rIndexFinger1_x",
"rIndexFinger1_y", "rIndexFinger2_x", "rIndexFinger2_y",
"rIndexFinger3_x", "rIndexFinger3_y", "rMiddleFinger0_x",
"rMiddleFinger0_y", "rMiddleFinger1_x", "rMiddleFinger1_y",
"rMiddleFinger2_x", "rMiddleFinger2_y", "rMiddleFinger3_x",
"rMiddleFinger3_y", "rRingFinger0_x", "rRingFinger0_y",
"rRingFinger1_x", "rRingFinger1_y", "rRingFinger2_x",
"rRingFinger2_y", "rRingFinger3_x", "rRingFinger3_y",
"rLittleFinger0_x", "rLittleFinger0_y", "rLittleFinger1_x",
"rLittleFinger1_y", "rLittleFinger2_x", "rLittleFinger2_y",
"rLittleFinger3_x", "rLittleFinger3_y"
])
frameElapsedTime = 1
totalTimeElapsed = 0
sampleIndex = 0
while True:
frameStart = time.time()
ret, frame = cam.read()
frame = cv2.flip(frame, 1)
datum.cvInputData = frame
openposeHandle.emplaceAndPop(op.VectorDatum([datum]))
if type(datum.poseKeypoints) != type(None):
if totalTimeElapsed > samplePeriod:
totalTimeElapsed -= samplePeriod
focusIndex = getFocusBodyIndex(datum.poseKeypoints,
len(datum.poseKeypoints))
body = datum.poseKeypoints[focusIndex]
leftHand = datum.handKeypoints[0][focusIndex]
rightHand = datum.handKeypoints[1][focusIndex]
cv2.imwrite(samplePath + "/%08d.jpg" % sampleIndex,
datum.cvOutputData)
csvWriter.writerow([
0, body[0][0], body[0][1], body[1][0], body[1][1],
body[2][0], body[2][1], body[3][0], body[3][1],
body[4][0], body[4][1], body[5][0], body[5][1],
body[6][0], body[6][1], body[7][0], body[7][1],
body[15][0], body[15][1], body[16][0], body[16][1],
body[17][0], body[17][1], body[18][0], body[18][1],
leftHand[0][0], leftHand[0][1], leftHand[1][0],
leftHand[1][1], leftHand[2][0], leftHand[2][1],
leftHand[3][0], leftHand[3][1], leftHand[4][0],
leftHand[4][1], leftHand[5][0], leftHand[5][1],
leftHand[6][0], leftHand[6][1], leftHand[7][0],
leftHand[7][1], leftHand[8][0], leftHand[8][1],
leftHand[9][0], leftHand[9][1], leftHand[10][0],
leftHand[10][1], leftHand[11][0], leftHand[11][1],
leftHand[12][0], leftHand[12][1], leftHand[13][0],
leftHand[13][1], leftHand[14][0], leftHand[14][1],
leftHand[15][0], leftHand[15][1], leftHand[16][0],
leftHand[16][1], leftHand[17][0], leftHand[17][1],
leftHand[18][0], leftHand[18][1], leftHand[19][0],
leftHand[19][1], leftHand[20][0], leftHand[20][1],
rightHand[0], rightHand[0][1], rightHand[1][0],
rightHand[1][1], rightHand[2][0], rightHand[2][1],
rightHand[3][0], rightHand[3][1], rightHand[4][0],
rightHand[4][1], rightHand[5][0], rightHand[5][1],
rightHand[6][0], rightHand[6][1], rightHand[7][0],
rightHand[7][1], rightHand[8][0], rightHand[8][1],
rightHand[9][0], rightHand[9][1], rightHand[10][0],
rightHand[10][1], rightHand[11][0], rightHand[11][1],
rightHand[12][0], rightHand[12][1], rightHand[13][0],
rightHand[13][1], rightHand[14][0], rightHand[14][1],
rightHand[15][0], rightHand[15][1], rightHand[16][0],
rightHand[16][1], rightHand[17][0], rightHand[17][1],
rightHand[18][0], rightHand[18][1], rightHand[19][0],
rightHand[19][1], rightHand[20][0], rightHand[20][1]
])
print("sample %08d saved." % sampleIndex)
sampleIndex += 1
datum.cvOutputData = cv2.putText(
datum.cvOutputData, "FPS: %.02f Target FPS: %.02f" %
(1.0 / frameElapsedTime, frameRate), (2, 34),
cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0))
datum.cvOutputData = cv2.putText(
datum.cvOutputData, "FPS: %.02f Target FPS: %.02f" %
(1.0 / frameElapsedTime, frameRate), (0, 32),
cv2.FONT_HERSHEY_DUPLEX, 1, (255, 255, 255))
cv2.imshow(windowName, datum.cvOutputData)
frameEnd = time.time()
remainTime = (1.0 / frameRate) - (frameEnd - frameStart)
if remainTime > 0:
key = cv2.waitKey(int(remainTime * 1000) & 0xffffffff)
else:
key = cv2.waitKey(1)
frameEnd = time.time()
frameElapsedTime = frameEnd - frameStart
totalTimeElapsed += frameElapsedTime
if cv2.getWindowProperty(windowName,
cv2.WND_PROP_ASPECT_RATIO) == -1:
break
cam.release()
