Only cast rays that intersect with the bbox

This commit is contained in:
Michel Breyer 2021-08-16 15:33:52 +02:00
parent 6da198a5bf
commit 4b4d54240b
3 changed files with 60 additions and 17 deletions

View File

@ -1,3 +1,4 @@
import itertools
import numpy as np import numpy as np
import active_grasp.msg import active_grasp.msg
@ -9,6 +10,10 @@ class AABBox:
self.min = bbox_min self.min = bbox_min
self.max = bbox_max self.max = bbox_max
@property
def corners(self):
return list(itertools.product(*np.vstack((self.min, self.max)).T))
def is_inside(self, p): def is_inside(self, p):
return np.all(p > self.min) and np.all(p < self.max) return np.all(p > self.min) and np.all(p < self.max)

View File

@ -1,11 +1,22 @@
import itertools import itertools
import numpy as np import numpy as np
from robot_helpers.perception import CameraIntrinsic
from robot_helpers.spatial import Transform
import rospy import rospy
from .policy import BasePolicy from .policy import BasePolicy
from vgn.utils import look_at, spherical_to_cartesian from vgn.utils import look_at, spherical_to_cartesian
class Ray:
def __init__(self, origin, direction):
self.o = origin
self.d = direction
def __call__(self, t):
return self.o + self.d * t
class NextBestView(BasePolicy): class NextBestView(BasePolicy):
def __init__(self, rate, filter_grasps): def __init__(self, rate, filter_grasps):
super().__init__(rate, filter_grasps) super().__init__(rate, filter_grasps)
@ -19,6 +30,7 @@ class NextBestView(BasePolicy):
# Generate viewpoints # Generate viewpoints
views = self.generate_viewpoints() views = self.generate_viewpoints()
self.visualizer.views(views)
# Evaluate viewpoints # Evaluate viewpoints
gains = [self.compute_ig(v) for v in views] gains = [self.compute_ig(v) for v in views]
@ -47,18 +59,25 @@ class NextBestView(BasePolicy):
return views return views
def compute_ig(self, view, downsample=20): def compute_ig(self, view, downsample=20):
res_x = int(self.intrinsic.width / downsample)
res_y = int(self.intrinsic.height / downsample)
fx = self.intrinsic.fx / downsample fx = self.intrinsic.fx / downsample
fy = self.intrinsic.fy / downsample fy = self.intrinsic.fy / downsample
cx = self.intrinsic.cx / downsample cx = self.intrinsic.cx / downsample
cy = self.intrinsic.cy / downsample cy = self.intrinsic.cy / downsample
for x in range(res_x): T_cam_base = view.inv()
for y in range(res_y): corners = np.array([T_cam_base.apply(p) for p in self.bbox.corners]).T
d = np.r_[(x - cx) / fx, (y - cy) / fy, 1.0] u = (fx * corners[0] / corners[2] + cx).round().astype(int)
d = d / np.linalg.norm(d) v = (fy * corners[1] / corners[2] + cy).round().astype(int)
d = view.rotation.apply(d) u_min, u_max = u.min(), u.max()
v_min, v_max = v.min(), v.max()
for u in range(u_min, u_max):
for v in range(v_min, v_max):
direction = np.r_[(u - cx) / fx, (v - cy) / fy, 1.0]
direction = direction / np.linalg.norm(direction)
direction = view.rotation.apply(direction)
ray = Ray(view.translation, direction)
return 1.0 return 1.0
def compute_cost(self, view): def compute_cost(self, view):

View File

@ -13,6 +13,7 @@ class Visualizer:
self.marker_pub = rospy.Publisher(topic, MarkerArray, queue_size=1) self.marker_pub = rospy.Publisher(topic, MarkerArray, queue_size=1)
self.scene_cloud_pub = rospy.Publisher("scene_cloud", PointCloud2, queue_size=1) self.scene_cloud_pub = rospy.Publisher("scene_cloud", PointCloud2, queue_size=1)
self.quality_pub = rospy.Publisher("quality", PointCloud2, queue_size=1) self.quality_pub = rospy.Publisher("quality", PointCloud2, queue_size=1)
self.views_pub = rospy.Publisher("views", PoseArray, queue_size=1)
self.grasps_pub = rospy.Publisher("grasps", PoseArray, queue_size=1) self.grasps_pub = rospy.Publisher("grasps", PoseArray, queue_size=1)
def clear(self): def clear(self):
@ -24,6 +25,9 @@ class Visualizer:
msg.header.frame_id = self.frame msg.header.frame_id = self.frame
self.grasps_pub.publish(msg) self.grasps_pub.publish(msg)
def draw(self, markers):
self.marker_pub.publish(MarkerArray(markers=markers))
def bbox(self, bbox): def bbox(self, bbox):
pose = Transform.translation((bbox.min + bbox.max) / 2.0) pose = Transform.translation((bbox.min + bbox.max) / 2.0)
scale = bbox.max - bbox.min scale = bbox.max - bbox.min
@ -31,6 +35,24 @@ class Visualizer:
marker = create_cube_marker(self.frame, pose, scale, color, ns="bbox") marker = create_cube_marker(self.frame, pose, scale, color, ns="bbox")
self.draw([marker]) self.draw([marker])
def grasps(self, grasps):
msg = PoseArray()
msg.header.frame_id = self.frame
msg.poses = [to_pose_msg(grasp.pose) for grasp in grasps]
self.grasps_pub.publish(msg)
def lines(self, lines):
marker = create_line_list_marker(
self.frame,
Transform.identity(),
[0.005, 0.0, 0.0],
[1.0, 0.0, 0.0],
lines,
"rays",
0,
)
self.draw([marker])
def path(self, poses): def path(self, poses):
color = np.r_[31, 119, 180] / 255.0 color = np.r_[31, 119, 180] / 255.0
points = [p.translation for p in poses] points = [p.translation for p in poses]
@ -54,20 +76,17 @@ class Visualizer:
) )
self.draw([spheres, lines]) self.draw([spheres, lines])
def draw(self, markers):
self.marker_pub.publish(MarkerArray(markers=markers))
def scene_cloud(self, frame, cloud):
msg = to_cloud_msg(frame, np.asarray(cloud.points))
self.scene_cloud_pub.publish(msg)
def quality(self, frame, voxel_size, quality): def quality(self, frame, voxel_size, quality):
points, values = grid_to_map_cloud(voxel_size, quality, threshold=0.8) points, values = grid_to_map_cloud(voxel_size, quality, threshold=0.8)
msg = to_cloud_msg(frame, points, intensities=values) msg = to_cloud_msg(frame, points, intensities=values)
self.quality_pub.publish(msg) self.quality_pub.publish(msg)
def grasps(self, grasps): def scene_cloud(self, frame, cloud):
msg = to_cloud_msg(frame, np.asarray(cloud.points))
self.scene_cloud_pub.publish(msg)
def views(self, views):
msg = PoseArray() msg = PoseArray()
msg.header.frame_id = self.frame msg.header.frame_id = self.frame
msg.poses = [to_pose_msg(grasp.pose) for grasp in grasps] msg.poses = [to_pose_msg(view) for view in views]
self.grasps_pub.publish(msg) self.views_pub.publish(msg)