width_of = 1080
height_of = 1080
for cam_idx, kps in enumerate(keypoints[subject][action]):
# Normalize camera frame
cam = dataset.cameras()[subject][cam_idx]
kps[..., :2] = normalize_screen_coordinates(kps[..., :2], w=width_of, h=height_of)
keypoints[subject][action][cam_idx] = kps
subjects_train = args.subjects_train.split(',')
subjects_semi = [] if not args.subjects_unlabeled else args.subjects_unlabeled.split(',')
subjects_test = args.subjects_test.split(',')
semi_supervised = len(subjects_semi) > 0
if semi_supervised and not dataset.supports_semi_supervised():
raise RuntimeError('Semi-supervised training is not implemented for this dataset')
def fetch(subjects, action_filter=None, subset=1, parse_3d_poses=True):
out_poses_3d = []
out_poses_2d = []
out_camera_params = []
subject = 'S1'
action = 'Directions'
poses_2d = keypoints[subject][action]
for i in range(len(poses_2d)): # Iterate across cameras
out_poses_2d.append(poses_2d[i])
for action in keypoints[subject]:
for cam_idx, kps in enumerate(keypoints[subject][action]):
# Normalize camera frame
cam = dataset.cameras()[subject][cam_idx]
kps[..., :2] = normalize_screen_coordinates(kps[..., :2], w=cam['res_w'], h=cam['res_h'])
keypoints[subject][action][cam_idx] = kps
subjects_train = args.subjects_train.split(',')
subjects_semi = [] if not args.subjects_unlabeled else args.subjects_unlabeled.split(',')
if not args.render:
subjects_test = args.subjects_test.split(',')
else:
subjects_test = [args.viz_subject]
semi_supervised = len(subjects_semi) > 0
if semi_supervised and not dataset.supports_semi_supervised():
raise RuntimeError('Semi-supervised training is not implemented for this dataset')
def fetch(subjects, action_filter=None, subset=1, parse_3d_poses=True):
out_poses_3d = []
out_poses_2d = []
out_camera_params = []
for subject in subjects:
for action in keypoints[subject].keys():
if action_filter is not None:
found = False
for a in action_filter:
if action.startswith(a):
found = True
break
if not found:
def the_main_kaboose(args):
print(args)
try:
# Create checkpoint directory if it does not exist
os.makedirs(args.checkpoint)
except OSError as e:
if e.errno != errno.EEXIST:
raise RuntimeError('Unable to create checkpoint directory:',
args.checkpoint)
print('Loading dataset...')
dataset_path = 'data/data_3d_' + args.dataset + '.npz'
if args.dataset == 'h36m':
from common.h36m_dataset import Human36mDataset
dataset = Human36mDataset(dataset_path)
elif args.dataset.startswith('humaneva'):
from common.humaneva_dataset import HumanEvaDataset
dataset = HumanEvaDataset(dataset_path)
elif args.dataset.startswith('custom'):
from common.custom_dataset import CustomDataset
dataset = CustomDataset('data/data_2d_' + args.dataset + '_' +
args.keypoints + '.npz')
else:
raise KeyError('Invalid dataset')
print('Preparing data...')
for subject in dataset.subjects():
for action in dataset[subject].keys():
anim = dataset[subject][action]
# this only works when training.
if 'positions' in anim:
positions_3d = []
for cam in anim['cameras']:
pos_3d = world_to_camera(anim['positions'],
R=cam['orientation'],
t=cam['translation'])
pos_3d[:,
1:] -= pos_3d[:, :
1] # Remove global offset, but keep trajectory in first position
positions_3d.append(pos_3d)
anim['positions_3d'] = positions_3d
print('Loading 2D detections...')
keypoints = np.load('data/data_2d_' + args.dataset + '_' + args.keypoints +
'.npz',
allow_pickle=True)
keypoints_metadata = keypoints['metadata'].item()
keypoints_symmetry = keypoints_metadata['keypoints_symmetry']
kps_left, kps_right = list(keypoints_symmetry[0]), list(
keypoints_symmetry[1])
joints_left, joints_right = list(dataset.skeleton().joints_left()), list(
dataset.skeleton().joints_right())
keypoints = keypoints['positions_2d'].item()
# THIS IS ABOUT TRAINING. ignore pls.
for subject in dataset.subjects():
assert subject in keypoints, 'Subject {} is missing from the 2D detections dataset'.format(
subject)
for action in dataset[subject].keys():
assert action in keypoints[
subject], 'Action {} of subject {} is missing from the 2D detections dataset'.format(
action, subject)
if 'positions_3d' not in dataset[subject][action]:
continue
for cam_idx in range(len(keypoints[subject][action])):
# We check for >= instead of == because some videos in H3.6M contain extra frames
mocap_length = dataset[subject][action]['positions_3d'][
cam_idx].shape[0]
assert keypoints[subject][action][cam_idx].shape[
0] >= mocap_length
if keypoints[subject][action][cam_idx].shape[0] > mocap_length:
# Shorten sequence
keypoints[subject][action][cam_idx] = keypoints[subject][
action][cam_idx][:mocap_length]
assert len(keypoints[subject][action]) == len(
dataset[subject][action]['positions_3d'])
# normalize camera frame?
for subject in keypoints.keys():
for action in keypoints[subject]:
for cam_idx, kps in enumerate(keypoints[subject][action]):
# Normalize camera frame
cam = dataset.cameras()[subject][cam_idx]
kps[..., :2] = normalize_screen_coordinates(kps[..., :2],
w=cam['res_w'],
h=cam['res_h'])
keypoints[subject][action][cam_idx] = kps
subjects_train = args.subjects_train.split(',')
subjects_semi = [] if not args.subjects_unlabeled else args.subjects_unlabeled.split(
',')
if not args.render:
subjects_test = args.subjects_test.split(',')
else:
subjects_test = [args.viz_subject]
semi_supervised = len(subjects_semi) > 0
if semi_supervised and not dataset.supports_semi_supervised():
raise RuntimeError(
'Semi-supervised training is not implemented for this dataset')
def fetch(subjects, action_filter=None, subset=1, parse_3d_poses=True):
out_poses_3d = []
out_poses_2d = []
out_camera_params = []
for subject in subjects:
print("gonna check actions for subject " + subject)
for subject in subjects:
for action in keypoints[subject].keys():
if action_filter is not None:
found = False
for a in action_filter:
if action.startswith(a):
found = True
break
if not found:
continue
poses_2d = keypoints[subject][action]
for i in range(len(poses_2d)): # Iterate across cameras
out_poses_2d.append(poses_2d[i])
if subject in dataset.cameras():
cams = dataset.cameras()[subject]
assert len(cams) == len(poses_2d), 'Camera count mismatch'
for cam in cams:
if 'intrinsic' in cam:
out_camera_params.append(cam['intrinsic'])
if parse_3d_poses and 'positions_3d' in dataset[subject][
action]:
poses_3d = dataset[subject][action]['positions_3d']
assert len(poses_3d) == len(
poses_2d), 'Camera count mismatch'
for i in range(len(poses_3d)): # Iterate across cameras
out_poses_3d.append(poses_3d[i])
if len(out_camera_params) == 0:
out_camera_params = None
if len(out_poses_3d) == 0:
out_poses_3d = None
stride = args.downsample
if subset < 1:
for i in range(len(out_poses_2d)):
n_frames = int(
round(len(out_poses_2d[i]) // stride * subset) * stride)
start = deterministic_random(
0,
len(out_poses_2d[i]) - n_frames + 1,
str(len(out_poses_2d[i])))
out_poses_2d[i] = out_poses_2d[i][start:start +
n_frames:stride]
if out_poses_3d is not None:
out_poses_3d[i] = out_poses_3d[i][start:start +
n_frames:stride]
elif stride > 1:
# Downsample as requested
for i in range(len(out_poses_2d)):
out_poses_2d[i] = out_poses_2d[i][::stride]
if out_poses_3d is not None:
out_poses_3d[i] = out_poses_3d[i][::stride]
return out_camera_params, out_poses_3d, out_poses_2d
action_filter = None if args.actions == '*' else args.actions.split(',')
if action_filter is not None:
print('Selected actions:', action_filter)
# when you run inference, this returns None, None, and the keypoints array renamed as poses_valid_2d
cameras_valid, poses_valid, poses_valid_2d = fetch(subjects_test,
action_filter)
filter_widths = [int(x) for x in args.architecture.split(',')]
if not args.disable_optimizations and not args.dense and args.stride == 1:
# Use optimized model for single-frame predictions
shape_2 = poses_valid_2d[0].shape[-2]
shape_1 = poses_valid_2d[0].shape[-1]
numJoints = dataset.skeleton().num_joints()
model_pos_train = TemporalModelOptimized1f(shape_2,
shape_1,
numJoints,
filter_widths=filter_widths,
causal=args.causal,
dropout=args.dropout,
channels=args.channels)
else:
# When incompatible settings are detected (stride > 1, dense filters, or disabled optimization) fall back to normal model
model_pos_train = TemporalModel(poses_valid_2d[0].shape[-2],
poses_valid_2d[0].shape[-1],
dataset.skeleton().num_joints(),
filter_widths=filter_widths,
causal=args.causal,
dropout=args.dropout,
channels=args.channels,
dense=args.dense)
model_pos = TemporalModel(poses_valid_2d[0].shape[-2],
poses_valid_2d[0].shape[-1],
dataset.skeleton().num_joints(),
filter_widths=filter_widths,
causal=args.causal,
dropout=args.dropout,
channels=args.channels,
dense=args.dense)
receptive_field = model_pos.receptive_field()
print('INFO: Receptive field: {} frames'.format(receptive_field))
pad = (receptive_field - 1) // 2 # Padding on each side
if args.causal:
print('INFO: Using causal convolutions')
causal_shift = pad
else:
causal_shift = 0
model_params = 0
for parameter in model_pos.parameters():
model_params += parameter.numel()
print('INFO: Trainable parameter count:', model_params)
if torch.cuda.is_available():
model_pos = model_pos.cuda()
model_pos_train = model_pos_train.cuda()
if args.resume or args.evaluate:
chk_filename = os.path.join(
args.checkpoint, args.resume if args.resume else args.evaluate)
print('Loading checkpoint', chk_filename)
checkpoint = torch.load(chk_filename,
map_location=lambda storage, loc: storage)
print('This model was trained for {} epochs'.format(
checkpoint['epoch']))
model_pos_train.load_state_dict(checkpoint['model_pos'])
model_pos.load_state_dict(checkpoint['model_pos'])
if args.evaluate and 'model_traj' in checkpoint:
# Load trajectory model if it contained in the checkpoint (e.g. for inference in the wild)
model_traj = TemporalModel(poses_valid_2d[0].shape[-2],
poses_valid_2d[0].shape[-1],
1,
filter_widths=filter_widths,
causal=args.causal,
dropout=args.dropout,
channels=args.channels,
dense=args.dense)
if torch.cuda.is_available():
model_traj = model_traj.cuda()
model_traj.load_state_dict(checkpoint['model_traj'])
else:
model_traj = None
test_generator = UnchunkedGenerator(cameras_valid,
poses_valid,
poses_valid_2d,
pad=pad,
causal_shift=causal_shift,
augment=False,
kps_left=kps_left,
kps_right=kps_right,
joints_left=joints_left,
joints_right=joints_right)
print('INFO: Testing on {} frames'.format(test_generator.num_frames()))
# Evaluate
def evaluate(eval_generator,
action=None,
return_predictions=False,
use_trajectory_model=False):
epoch_loss_3d_pos = 0
epoch_loss_3d_pos_procrustes = 0
epoch_loss_3d_pos_scale = 0
epoch_loss_3d_vel = 0
with torch.no_grad():
if not use_trajectory_model:
model_pos.eval()
else:
model_traj.eval()
N = 0
for _, batch, batch_2d in eval_generator.next_epoch():
inputs_2d = torch.from_numpy(batch_2d.astype('float32'))
if torch.cuda.is_available():
inputs_2d = inputs_2d.cuda()
# Positional model
if not use_trajectory_model:
predicted_3d_pos = model_pos(inputs_2d)
else:
predicted_3d_pos = model_traj(inputs_2d)
# Test-time augmentation (if enabled)
if eval_generator.augment_enabled():
# Undo flipping and take average with non-flipped version
predicted_3d_pos[1, :, :, 0] *= -1
if not use_trajectory_model:
predicted_3d_pos[1, :, joints_left +
joints_right] = predicted_3d_pos[
1, :, joints_right + joints_left]
predicted_3d_pos = torch.mean(predicted_3d_pos,
dim=0,
keepdim=True)
if return_predictions:
return predicted_3d_pos.squeeze(0).cpu().numpy()
inputs_3d = torch.from_numpy(batch.astype('float32'))
if torch.cuda.is_available():
inputs_3d = inputs_3d.cuda()
inputs_3d[:, :, 0] = 0
if eval_generator.augment_enabled():
inputs_3d = inputs_3d[:1]
error = mpjpe(predicted_3d_pos, inputs_3d)
epoch_loss_3d_pos_scale += inputs_3d.shape[
0] * inputs_3d.shape[1] * n_mpjpe(predicted_3d_pos,
inputs_3d).item()
epoch_loss_3d_pos += inputs_3d.shape[0] * inputs_3d.shape[
1] * error.item()
N += inputs_3d.shape[0] * inputs_3d.shape[1]
inputs = inputs_3d.cpu().numpy().reshape(
-1, inputs_3d.shape[-2], inputs_3d.shape[-1])
predicted_3d_pos = predicted_3d_pos.cpu().numpy().reshape(
-1, inputs_3d.shape[-2], inputs_3d.shape[-1])
epoch_loss_3d_pos_procrustes += inputs_3d.shape[
0] * inputs_3d.shape[1] * p_mpjpe(predicted_3d_pos, inputs)
# Compute velocity error
epoch_loss_3d_vel += inputs_3d.shape[0] * inputs_3d.shape[
1] * mean_velocity_error(predicted_3d_pos, inputs)
if action is None:
print('----------')
else:
print('----' + action + '----')
e1 = (epoch_loss_3d_pos / N) * 1000
e2 = (epoch_loss_3d_pos_procrustes / N) * 1000
e3 = (epoch_loss_3d_pos_scale / N) * 1000
ev = (epoch_loss_3d_vel / N) * 1000
print('Test time augmentation:', eval_generator.augment_enabled())
print('Protocol #1 Error (MPJPE):', e1, 'mm')
print('Protocol #2 Error (P-MPJPE):', e2, 'mm')
print('Protocol #3 Error (N-MPJPE):', e3, 'mm')
print('Velocity Error (MPJVE):', ev, 'mm')
print('----------')
return e1, e2, e3, ev
if args.render:
print('Rendering...')
input_keypoints = keypoints[args.viz_subject][args.viz_action][
args.viz_camera].copy()
ground_truth = None
if args.viz_subject in dataset.subjects(
) and args.viz_action in dataset[args.viz_subject]:
if 'positions_3d' in dataset[args.viz_subject][args.viz_action]:
ground_truth = dataset[args.viz_subject][
args.viz_action]['positions_3d'][args.viz_camera].copy()
if ground_truth is None:
print(
'INFO: this action is unlabeled. Ground truth will not be rendered.'
)
gen = UnchunkedGenerator(None,
None, [input_keypoints],
pad=pad,
causal_shift=causal_shift,
augment=args.test_time_augmentation,
kps_left=kps_left,
kps_right=kps_right,
joints_left=joints_left,
joints_right=joints_right)
prediction = evaluate(gen, return_predictions=True)
if model_traj is not None and ground_truth is None:
prediction_traj = evaluate(gen,
return_predictions=True,
use_trajectory_model=True)
prediction += prediction_traj
if args.viz_export is not None:
print('Exporting joint positions to', args.viz_export)
# Predictions are in camera space
np.save(args.viz_export, prediction)
if args.viz_output is not None:
if ground_truth is not None:
# Reapply trajectory
trajectory = ground_truth[:, :1]
ground_truth[:, 1:] += trajectory
prediction += trajectory
# Invert camera transformation
cam = dataset.cameras()[args.viz_subject][args.viz_camera]
if ground_truth is not None:
prediction = camera_to_world(prediction,
R=cam['orientation'],
t=cam['translation'])
ground_truth = camera_to_world(ground_truth,
R=cam['orientation'],
t=cam['translation'])
else:
# If the ground truth is not available, take the camera extrinsic params from a random subject.
# They are almost the same, and anyway, we only need this for visualization purposes.
for subject in dataset.cameras():
if 'orientation' in dataset.cameras()[subject][
args.viz_camera]:
rot = dataset.cameras()[subject][
args.viz_camera]['orientation']
break
prediction = camera_to_world(prediction, R=rot, t=0)
# We don't have the trajectory, but at least we can rebase the height
prediction[:, :, 2] -= np.min(prediction[:, :, 2])
anim_output = {'Reconstruction': prediction}
if ground_truth is not None and not args.viz_no_ground_truth:
anim_output['Ground truth'] = ground_truth
input_keypoints = image_coordinates(input_keypoints[..., :2],
w=cam['res_w'],
h=cam['res_h'])
print("Writing to json")
import json
# format the data in the same format as mediapipe, so we can load it in unity with the same script
# we need a list (frames) of lists of 3d landmarks.
# but prediction[] only has 17 landmarks, and we need 25 in our unity script
unity_landmarks = prediction.tolist()
with open(args.output_json, "w") as json_file:
json.dump(unity_landmarks, json_file)
if args.rendervideo == "yes":
from common.visualization import render_animation
render_animation(input_keypoints,
keypoints_metadata,
anim_output,
dataset.skeleton(),
dataset.fps(),
args.viz_bitrate,
cam['azimuth'],
args.viz_output,
limit=args.viz_limit,
downsample=args.viz_downsample,
size=args.viz_size,
input_video_path=args.viz_video,
viewport=(cam['res_w'], cam['res_h']),
input_video_skip=args.viz_skip)
