def refine_poses(self,
keypoint_thresh=10,
score_thresh=0.5,
neck_thresh=0.59,
margin=0.0):
W, H = 104.73, 67.74
for i, basename in enumerate(tqdm(self.frame_basenames)):
poses = self.poses[basename]
# remove the poses with few keypoints or they
keep = []
for ii in range(len(poses)):
keypoints = poses[ii]
valid = (keypoints[:, 2] > 0.).nonzero()[0]
score = np.sum(keypoints[valid, 2])
if len(
valid
) > keypoint_thresh and score > score_thresh and keypoints[
1, 2] > neck_thresh:
keep.append(ii)
poses = [poses[ii] for ii in keep]
root_part = 1
root_box = []
for ii in range(len(poses)):
root_tmp = poses[ii][root_part, :]
valid_keypoints = (poses[ii][:, 2] > 0).nonzero()
root_box.append([
root_tmp[0] - 10, root_tmp[1] - 10, root_tmp[0] + 10,
root_tmp[1] + 10,
np.sum(poses[ii][valid_keypoints, 2])
])
root_box = np.array(root_box)
# Perform Neck NMS
if len(root_box.shape) == 1:
root_box = root_box[None, :]
keep2 = [0]
else:
keep2 = nms(root_box.astype(np.float32), 0.1)
poses = [poses[ii] for ii in keep2 if ii < len(poses)]
# Remove poses outside of field
keep3 = []
cam_mat = self.calib[basename]
cam = cam_utils.Camera(basename, cam_mat['A'], cam_mat['R'],
cam_mat['T'], self.shape[0], self.shape[1])
for ii in range(len(poses)):
kp3 = misc_utils.lift_keypoints_in_3d(cam, poses[ii])
if (-W / 2. - margin) <= kp3[1, 0] <= (W / 2. + margin) and (
-H / 2. - margin) <= kp3[1, 2] <= (H / 2. + margin):
keep3.append(ii)
poses = [poses[ii] for ii in keep3]
self.poses[basename] = poses
fig = plt.figure()
ax = fig.add_subplot(111)
for i in tqdm(range(len(new_tracklets))):
neck_pos = []
for j in range(len(new_tracklets[i])):
frame_index = new_tracklets[i][j].frame_index
basename = db.frame_basenames[frame_index]
cam_data = db.calib[basename]
cam = cam_utils.Camera(basename, cam_data['A'], cam_data['R'], cam_data['T'], db.shape[0], db.shape[1])
kp_3d = misc_utils.lift_keypoints_in_3d(cam, new_tracklets[i][j].keypoints)
neck_pos.append(kp_3d[1, :])
neck_pos = np.array(neck_pos)
# Smooth trajectory
smoothed_positions = smooth_trajectory(new_tracklets[i], neck_pos)
for j in range(len(new_tracklets[i])):
data_out[new_tracklets[i][j].frame].append({'mesh': new_tracklets[i][j].mesh_name, 'x': smoothed_positions[0, j],
'y': smoothed_positions[1, j], 'z': smoothed_positions[2, j]})
ax.plot(smoothed_positions[0, :], smoothed_positions[2, :], 'o')
plt.show()
with open(join(db.path_to_dataset, 'players', 'metadata', 'position.json'), 'w') as outfile:
json.dump(data_out, outfile)
basename = db.frame_basenames[sel_frame]
poses = db.poses[basename]
mask = db.get_instances_from_detectron(sel_frame, is_bool=True)
cam_mat = db.calib[basename]
cam = cam_utils.Camera(basename, cam_mat['A'], cam_mat['R'], cam_mat['T'],
db.shape[0], db.shape[1])
skeleton_buffer = seg_utils.get_instance_skeleton_buffer(
db.shape[0], db.shape[1], poses)
h, w = img.shape[:2]
for i in range(len(poses)):
valid = poses[i][:, 2] > 0
kp3 = misc_utils.lift_keypoints_in_3d(cam, poses[i][valid, :], pad=0)
center3d = np.mean(kp3, axis=0)
# Most of keypoitns are in the upper body so the center of the mass is closer to neck
center3d[1] -= 0.25
_, center_depth = cam.project(np.array([center3d]))
bbox = misc_utils.get_box_from_3d_shpere(cam, center3d)
x1, y1, x2, y2 = bbox[:4]
x1 -= margin
y1 -= margin
x2 += margin
y2 += margin
x1, x2, y1, y2 = max(x1, 0), min(w, x2), max(y1, 0), min(h, y2)