def process_frame(txt_file, frame, body=True, hands=False):
canvas = copy.deepcopy(frame)
if body:
candidate, subset = body_estimation(frame)
if (is_data_valid(subset)):
body_points = get_18_body_points(candidate, subset)
list_to_file(txt_file, body_points)
else:
canvas = copy.deepcopy(frame)
canvas = util.draw_bodypose(canvas, candidate, subset)
# plt.imshow(canvas[:, :, [2, 1, 0]])
# plt.axis('off')
# plt.show()
print(subset)
if hands:
hands_list = util.handDetect(candidate, subset, frame)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
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)
canvas = util.draw_handpose(canvas, all_hand_peaks)
return canvas
def main(oriImg):
shape0 = oriImg.shape
candidate, subset = body_estimation(oriImg)
canvas = copy.deepcopy(oriImg)
shape1 = canvas.shape
canvas = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)
all_hand_peaks = []
shape2 = canvas.shape
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)
shape3 = canvas.shape
canvas = util.draw_handpose(canvas, all_hand_peaks)
shape4 = canvas.shape
cv2.imwrite("test.png", canvas)
canvas = cv2.resize(canvas, (shape0[1],shape0[0]))
# raise AttributeError(shape0, shape1, shape2, shape3, shape4)
return canvas[:, :, [2, 1, 0]]
def detect_keypoint(test_image, is_vis):
body_estimation = Body('model/body_pose_model.pth')
hand_estimation = Hand('model/hand_pose_model.pth')
oriImg = cv2.imread(test_image) # B,G,R order
# detect body
# subset: n*20 array, n is the human_number in the index, 0-17 is the index in candidate, 18 is the total score, 19 is the total parts
# candidate: m*4, m is the keypoint number in the image, [x, y, confidence, id]
candidate, subset = body_estimation(
oriImg
) # candidate: output the keypoints([25, 4]), x, y, score, keypoint_index
canvas = copy.deepcopy(oriImg)
canvas, bodypoints = util.draw_bodypose(canvas, candidate, subset)
# detect hand
hands_list = util.handDetect(candidate, subset, oriImg)
all_hand_peaks = []
hand_personid_isleft = []
for x, y, w, is_left, person_id 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)
hand_personid_isleft.append([person_id, is_left])
# all_hand_peaks: [p, 21, 2] p is the hand number in the image
# hand_personid_isleft: [p, 2] is_isleft, person_id
all_hand_peaks = np.asarray(all_hand_peaks)
hand_personid_isleft = np.asarray(hand_personid_isleft)
canvas = util.draw_handpose(canvas, all_hand_peaks)
if is_vis:
plt.imshow(canvas[:, :, [2, 1, 0]])
plt.axis('off')
plt.show()
return bodypoints, all_hand_peaks, hand_personid_isleft
def process_frame(frame, body=True, hands=True):
canvas = copy.deepcopy(frame)
if body:
candidate, subset = body_estimation(frame)
canvas = util.draw_bodypose(canvas, candidate, subset)
if hands:
hands_list = util.handDetect(candidate, subset, frame)
all_hand_peaks = []
for x, y, w, is_left in hands_list:
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)
canvas = util.draw_handpose(canvas, all_hand_peaks)
return canvas
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'), 10,
(frame_width, frame_height))
frame_count = 0
start = time.time()
while True:
isTrue, frame = capture.read()
if isTrue == True:
frame_count += 1
print('Processing %dth frame...' % frame_count)
candidate, subset = body_estimation(frame)
canvas = copy.deepcopy(frame)
canvas = util.draw_bodypose(canvas, candidate, subset)
util.toTempList(candidate)
# Write the frame into the file 'output.avi'
out.write(canvas)
# Display the resulting frame
cv2.imshow('Preview', canvas)
# Press Q on keyboard to stop recording
if cv2.waitKey(1) & 0xFF == ord('d'):
break
# Break the loop
else:
elif not found and k < 17:
row = -1 * np.ones(20)
row[indexA] = partAs[i]
row[indexB] = partBs[i]
row[-1] = 2
row[-2] = sum(
candidate[connection_all[k][i, :2].astype(int),
2]) + connection_all[k][i][2]
subset = np.vstack([subset, row])
# delete some rows of subset which has few parts occur
deleteIdx = []
for i in range(len(subset)):
if subset[i][-1] < 4 or subset[i][-2] / subset[i][-1] < 0.4:
deleteIdx.append(i)
subset = np.delete(subset, deleteIdx, axis=0)
# subset: n*20 array, 0-17 is the index in candidate, 18 is the total score, 19 is the total parts
# candidate: x, y, score, id
return candidate, subset, heatmap_list, heatmap_list_converted_list
if __name__ == "__main__":
body_estimation = Body('../model/body_pose_model.pth')
test_image = '../images/ski.jpg'
oriImg = cv2.imread(test_image) # B,G,R order
candidate, subset = body_estimation(oriImg)
canvas = util.draw_bodypose(oriImg, candidate, subset)
plt.imshow(canvas[:, :, [2, 1, 0]])
plt.show()
def process_one(video_input):
output_fname = './output-vid/' + video_input.split('/')[-1].split(
'.')[0] + '_openpose_inference_kp.avi'
print(output_fname)
if os.path.isfile(output_fname):
print("file alreay exist, return and process next")
return
cap = cv2.VideoCapture(video_input)
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
frames_per_second = cap.get(cv2.CAP_PROP_FPS)
# cap.set(3, 640)
# cap.set(4, 480)
output_file = cv2.VideoWriter(
filename=output_fname,
# some installation of opencv may not support x264 (due to its license),
# you can try other format (e.g. MPEG)
# apiPreference=1,
fourcc=cv2.VideoWriter_fourcc(*'XVID'),
# fourcc=cv2.VideoWriter_fourcc(*"x264"),
fps=float(frames_per_second),
frameSize=(width, height),
isColor=True)
# people = {}
# people['pose_keypoints_2d'] = {}
# cnt = 0
while True:
# cnt += 1
ret, oriImg = cap.read()
if oriImg is None:
break
candidate, subset = body_estimation(oriImg)
# save the 2d-pose results to JSON as the OpenPose output format
# people['pose_keypoints_2d'][cnt] = []
# for i in range(18):
# for n in range(len(subset)):
# index = int(subset[n][i])
# xyc = [-1, -1, -1] if index == -1 else candidate[index][0:3]
# people['pose_keypoints_2d'][cnt].extend(xyc)
#
# canvas = copy.deepcopy(oriImg)
# canvas = util.draw_bodypose(canvas, candidate, subset)
canvas = util.draw_bodypose(oriImg, candidate, subset)
# detect hand
# hands_list = util.handDetect(candidate, subset, oriImg)
#
# all_hand_peaks = []
# for x, y, w, is_left in hands_list:
# 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)
#
# canvas = util.draw_handpose(canvas, all_hand_peaks)
#
cv2.imshow('demo', canvas) # 一个窗口用以显示原视频
output_file.write(canvas)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
output_file.release()
cv2.destroyAllWindows()
test_image1 = 'images/aa.jpg' # 비교할 이미지 1
test_image2 = 'images/dragon2.jpg' # 비교할 이미지 2
oriImg1 = cv2.imread(test_image1) # 이미지 불러오기
candidate1, subset1 = body_estimation(
oriImg1) # candidate: x, y, score, id | subset: index
# x와 y가 관절의 위치, index는 관절의 이름이라고 생각하면 됨
# score는 단순하게 확실 정도라고 보면됨
# id는 무시해도 되고
oriImg2 = cv2.imread(test_image2) # 이미지2 불러오기
candidate2, subset2 = body_estimation(oriImg2)
canvas1 = copy.deepcopy(oriImg1) # 이미지 복사
canvas1 = util.draw_bodypose(canvas1, candidate1, subset1) # 화면에 예측한 스켈레톤을 보여줌
canvas2 = copy.deepcopy(oriImg2)
canvas2 = util.draw_bodypose(canvas2, candidate2, subset2)
mae = pose_compare.mae(candidate1, subset1, candidate2,
subset2) # 관절들의 위치를 통해 뼈의 각도를 알아내고
# 두이미지의 각도의 mean average error를 구한다
print("자세 오류: ", mae)
plt.subplot(121)
plt.imshow(canvas1[:, :, [2, 1, 0]])
plt.axis('off')
plt.subplot(122)
plt.imshow(canvas2[:, :, [2, 1, 0]])
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue Apr 28 19:17:47 2020
@author: joe
"""
from src import torch_openpose, util
import cv2
if __name__ == "__main__":
tp = torch_openpose.torch_openpose('body_25')
img = cv2.imread("images/timg.jpeg")
poses = tp(img)
img = util.draw_bodypose(img, poses, 'body_25')
cv2.imshow('v', img)
cv2.waitKey(0)
cv2.destroyAllWindows()
