File size: 8,202 Bytes
a199a9b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 |
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
from collections import deque
from dataclasses import dataclass
from pathlib import Path
from uuid import uuid4
import numpy as np
import rerun as rr
import rerun.blueprint as rrb
from loguru import logger
from nymeria.data_provider import NymeriaDataProvider
from PIL import Image
from projectaria_tools.core.sensor_data import ImageData
from projectaria_tools.core.sophus import SE3
from tqdm import tqdm
@dataclass(frozen=True)
class ViewerConfig:
output_rrd: Path = None
sample_fps: float = 10
rotate_rgb: bool = True
downsample_rgb: bool = True
jpeg_quality: int = 90
traj_tail_length: int = 100
ep_recording_head: str = "recording_head/2d"
ep_recording_observer: str = "recording_observer/2d"
point_radii: float = 0.008
line_radii: float = 0.008
skel_radii: float = 0.01
class NymeriaViewer(ViewerConfig):
palette: dict[str, list] = {
"recording_head": [255, 0, 0],
"recording_lwrist": [0, 255, 0],
"recording_rwrist": [0, 0, 255],
"recording_observer": [61, 0, 118],
"pointcloud": [128, 128, 128, 128],
"momentum": [218, 234, 134],
}
color_skeleton = np.array(
[
[127, 0, 255],
[105, 34, 254],
[81, 71, 252],
[59, 103, 249],
[35, 136, 244],
[11, 167, 238],
[10, 191, 232],
[34, 214, 223],
[58, 232, 214],
[80, 244, 204],
[104, 252, 192],
[128, 254, 179],
[150, 252, 167],
[174, 244, 152],
[196, 232, 138],
[220, 214, 122],
[244, 191, 105],
[255, 167, 89],
[255, 136, 71],
[255, 103, 53],
[255, 71, 36],
[255, 34, 17],
]
)
def __init__(self, **kwargs) -> None:
super().__init__(**kwargs)
blueprint = rrb.Horizontal(
rrb.Spatial3DView(name="3d"),
rrb.Vertical(
rrb.Spatial2DView(name="2d participant", origin=self.ep_recording_head),
rrb.Spatial2DView(
name="2d observer", origin=self.ep_recording_observer
),
),
)
rr.init(
"nymeria data viewer",
spawn=(self.output_rrd is None),
recording_id=uuid4(),
default_blueprint=blueprint,
)
if self.output_rrd is not None:
rr.save(self.output_rrd)
rr.log("world", rr.ViewCoordinates.RIGHT_HAND_Z_UP, static=True)
self._init_mesh: bool = False
self._epaths_3d: set[str] = set()
self._traj_deques: dict[str, deque] = {}
def __call__(self, nymeria_dp: NymeriaDataProvider):
# add static scene
self.__log_pointcloud(nymeria_dp)
self.__log_trajectory(nymeria_dp)
# add dynamic scene
t_ns_start, t_ns_end = nymeria_dp.timespan_ns
dt: int = int(1e9 / self.sample_fps)
for idx, t_ns in tqdm(enumerate(range(t_ns_start, t_ns_end, dt))):
rr.set_time_sequence("frames", idx)
rr.set_time_nanos("timestamps_ns", t_ns)
self.__log_synced_video(t_ns, nymeria_dp)
self.__log_synced_poses(t_ns, nymeria_dp)
self.__set_viewpoint()
def __log_pointcloud(self, nymeria_dp: NymeriaDataProvider) -> None:
pointclouds = nymeria_dp.get_all_pointclouds()
for tag, pts in pointclouds.items():
logger.info(f"add point cloud {tag}")
cc = self.palette.get("pointcloud")
ep = f"world/semidense_pts/{tag}"
rr.log(
entity_path=ep,
entity=rr.Points3D(pts, colors=cc, radii=self.point_radii),
static=True,
)
self._epaths_3d.add(ep)
def __log_trajectory(self, nymeria_dp: NymeriaDataProvider) -> None:
trajs: dict[str, np.ndarray] = nymeria_dp.get_all_trajectories()
for tag, traj in trajs.items():
logger.info(f"add trajectory {tag}, {traj.shape=}")
ep = f"world/traj_full/{tag}"
rr.log(
ep,
rr.LineStrips3D(
traj[:, :3, 3], colors=self.palette.get(tag), radii=self.line_radii
),
static=True,
)
self._epaths_3d.add(ep)
def __log_synced_video(self, t_ns: int, nymeria_dp: NymeriaDataProvider) -> None:
images: dict[str, tuple] = nymeria_dp.get_synced_rgb_videos(t_ns)
for tag, data in images.items():
rgb: ImageData = data[0]
if self.downsample_rgb:
rgb = rgb.to_numpy_array()[::2, ::2, :]
rgb = Image.fromarray(rgb.astype(np.uint8))
if self.rotate_rgb:
rgb = rgb.rotate(-90)
if tag in self.ep_recording_head:
ep = self.ep_recording_head
elif tag in self.ep_recording_observer:
ep = self.ep_recording_observer
rr.log(
f"{ep}/214-1", rr.Image(rgb).compress(jpeg_quality=self.jpeg_quality)
)
def __log_synced_poses(self, t_ns: int, nymeria_dp: NymeriaDataProvider) -> None:
poses: dict[str, any] = nymeria_dp.get_synced_poses(t_ns)
self._T_mv: SE3 = None
for tag, val in poses.items():
if "recording" in tag and self.traj_tail_length > 0:
traj = self._traj_deques.setdefault(tag, deque())
if self.traj_tail_length > 0 and len(traj) == self.traj_tail_length:
traj.popleft()
t = val.transform_world_device.translation()
traj.append(t.squeeze().tolist())
ep = f"world/traj_tail/{tag}"
rr.log(
ep,
rr.LineStrips3D(
traj, colors=self.palette.get(tag), radii=self.line_radii
),
)
self._epaths_3d.add(ep)
if tag == "xsens":
ep = "world/body/xsens_skel"
logger.debug(f"xsens skeleton {val.shape = }")
rr.log(
ep,
rr.LineStrips3D(
val, colors=self.color_skeleton, radii=self.skel_radii
),
static=False,
)
self._epaths_3d.add(ep)
if tag == "momentum":
ep = "world/body/momentum_mesh"
if self._init_mesh:
rr.log_components(ep, [rr.components.Position3DBatch(val)])
else:
faces = nymeria_dp.body_dp.momentum_template_mesh.faces
normals = nymeria_dp.body_dp.momentum_template_mesh.normals
rr.log(
ep,
rr.Mesh3D(
triangle_indices=faces,
vertex_positions=val,
vertex_normals=normals,
vertex_colors=self.palette.get(tag),
),
)
self._init_mesh = True
self._epaths_3d.add(ep)
if tag == "recording_head":
self._T_mv = val.transform_world_device
def __set_viewpoint(self, add_rotation: bool = False):
if self._T_mv is None:
return
t = self._T_mv.translation() * -1.0
Rz = np.eye(3)
if add_rotation:
R = self._T_mv.rotation().to_matrix()
psi = np.arctan2(R[1, 0], R[0, 0])
Rz[0:2, 0:2] = np.array(
[np.cos(psi), -np.sin(psi), np.sin(psi), np.cos(psi)]
).reshape(2, 2)
for ep in self._epaths_3d:
rr.log(
ep,
rr.Transform3D(translation=t, mat3x3=Rz),
static=False,
)
|