run demo_webcam_smooth with box pose

demo_webcam_smooth.py

@@ -13,6 +13,7 @@
 from FaceBoxes import FaceBoxes
 from TDDFA import TDDFA
 from utils.render import render
+from utils.pose import viz_pose
 # from utils.render_ctypes import render
 from utils.functions import cv_draw_landmark
 
@@ -99,6 +100,8 @@ def main(args):
  img_draw = cv_draw_landmark(queue_frame[n_pre], ver_ave, size=1)
  elif args.opt == '3d':
  img_draw = render(queue_frame[n_pre], [ver_ave], tddfa.tri, alpha=0.7)
+ elif args.opt == 'box_pose':
+ img_draw = viz_pose(queue_frame[n_pre], param_lst, [ver_ave], show_flag=False)
  else:
  raise ValueError(f'Unknown opt {args.opt}')
 
@@ -115,7 +118,7 @@ def main(args):
  parser = argparse.ArgumentParser(description='The smooth demo of webcam of 3DDFA_V2')
  parser.add_argument('-c', '--config', type=str, default='configs/mb1_120x120.yml')
  parser.add_argument('-m', '--mode', default='cpu', type=str, help='gpu or cpu mode')
- parser.add_argument('-o', '--opt', type=str, default='2d_sparse', choices=['2d_sparse', '2d_dense', '3d'])
+ parser.add_argument('-o', '--opt', type=str, default='2d_sparse', choices=['2d_sparse', '2d_dense', '3d', "box_pose"])
  parser.add_argument('-n_pre', default=1, type=int, help='the pre frames of smoothing')
  parser.add_argument('-n_next', default=1, type=int, help='the next frames of smoothing')
  parser.add_argument('--onnx', action='store_true', default=False)

readme.md

@@ -80,6 +80,10 @@ python3 demo_video_smooth.py -f examples/inputs/videos/214.avi --onnx
 
 # 4. running on webcam
 python3 demo_webcam_smooth.py --onnx
+
+# 5. running on webcam with box pose
+
+python3 demo_webcam_smooth.py --onnx --opt box_pose
 ```
 
 The implementation of tracking is simply by alignment. If the head pose > 90° or the motion is too fast, the alignment may fail. A threshold is used to trickly check the tracking state, but it is unstable.
@@ -104,6 +108,11 @@ Running on a video will give:
  <img src="docs/images/out.gif" alt="demo" width="512px">
 </p>
 
+Running on a webcam will give:
+
+<p align="center">
+ <img src="examples/inputs/box_pose.jpeg", width="512px">
+</p>
 More results or demos to see: [Hathaway](https://guojianzhu.com/assets/videos/hathaway_3ddfa_v2.mp4).
 
 <!-- Obviously, the eyes parts are not good. -->

utils/pose.py

@@ -130,12 +130,10 @@ def viz_pose(img, param_lst, ver_lst, show_flag=False, wfp=None):
  img = plot_pose_box(img, P, ver)
  # print(P[:, :3])
  print(f'yaw: {pose[0]:.1f}, pitch: {pose[1]:.1f}, roll: {pose[2]:.1f}')
-
+ # writing yaw, pitch, roll in the upper left of the display
+ img = cv2.putText(img, f"yaw: {pose[0]:.1f} pitch: {pose[1]:.1f} roll: {pose[2]:.1f}", (10, 20), 2, .8, (255, 0, 0), 2)
  if wfp is not None:
  cv2.imwrite(wfp, img)
  print(f'Save visualization result to {wfp}')
 
- if show_flag:
- plot_image(img)
-
  return img