hofee/nbv_reconstruction
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update new_num limit

Browse Source
This commit is contained in:
hofee 2024-10-05 15:36:38 -05:00
parent 1a3ae15130
commit e315fd99ee
2 changed files with 20 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
core/seq_dataset.py
View File

@ -74,24 +74,12 @@ class SeqNBVReconstructionDataset(BaseDataset):
max_coverage_rate = data_item_info["max_coverage_rate"] max_coverage_rate = data_item_info["max_coverage_rate"]
scene_name = data_item_info["scene_name"] scene_name = data_item_info["scene_name"]
first_cam_info = DataLoadUtil.load_cam_info(DataLoadUtil.get_path(self.root_dir, scene_name, first_frame_idx), binocular=True) first_cam_info = DataLoadUtil.load_cam_info(DataLoadUtil.get_path(self.root_dir, scene_name, first_frame_idx), binocular=True)
first_view_path = DataLoadUtil.get_path(self.root_dir, scene_name, first_frame_idx) first_view_path = DataLoadUtil.get_path(self.root_dir, scene_name, first_frame_idx)
first_left_cam_pose = first_cam_info["cam_to_world"] first_left_cam_pose = first_cam_info["cam_to_world"]
first_right_cam_pose = first_cam_info["cam_to_world_R"]
first_center_cam_pose = first_cam_info["cam_to_world_O"] first_center_cam_pose = first_cam_info["cam_to_world_O"]
if self.load_from_preprocess: first_target_point_cloud = DataLoadUtil.load_from_preprocessed_pts(first_view_path)
first_downsampled_target_point_cloud = DataLoadUtil.load_from_preprocessed_pts(first_view_path) first_pts_num = first_target_point_cloud.shape[0]
else: first_downsampled_target_point_cloud = PtsUtil.random_downsample_point_cloud(first_target_point_cloud, self.pts_num)
first_depth_L, first_depth_R = DataLoadUtil.load_depth(first_view_path, first_cam_info['near_plane'], first_cam_info['far_plane'], binocular=True)
first_point_cloud_L = DataLoadUtil.get_point_cloud(first_depth_L, first_cam_info['cam_intrinsic'], first_left_cam_pose)['points_world']
first_point_cloud_R = DataLoadUtil.get_point_cloud(first_depth_R, first_cam_info['cam_intrinsic'], first_right_cam_pose)['points_world']
first_point_cloud_L = PtsUtil.random_downsample_point_cloud(first_point_cloud_L, 65536)
first_point_cloud_R = PtsUtil.random_downsample_point_cloud(first_point_cloud_R, 65536)
first_overlap_points = PtsUtil.get_overlapping_points(first_point_cloud_L, first_point_cloud_R)
first_downsampled_target_point_cloud = PtsUtil.random_downsample_point_cloud(first_overlap_points, self.pts_num)
first_to_world_rot_6d = PoseUtil.matrix_to_rotation_6d_numpy(np.asarray(first_left_cam_pose[:3,:3])) first_to_world_rot_6d = PoseUtil.matrix_to_rotation_6d_numpy(np.asarray(first_left_cam_pose[:3,:3]))
first_to_world_trans = first_left_cam_pose[:3,3] first_to_world_trans = first_left_cam_pose[:3,3]
first_to_world_9d = np.concatenate([first_to_world_rot_6d, first_to_world_trans], axis=0) first_to_world_9d = np.concatenate([first_to_world_rot_6d, first_to_world_trans], axis=0)
@ -102,6 +90,9 @@ class SeqNBVReconstructionDataset(BaseDataset):
model_points_normals = DataLoadUtil.load_points_normals(self.root_dir, scene_name) model_points_normals = DataLoadUtil.load_points_normals(self.root_dir, scene_name)
data_item = { data_item = {
"first_pts_num": np.asarray(
first_pts_num, dtype=np.int32
),
"first_pts": np.asarray([first_downsampled_target_point_cloud],dtype=np.float32), "first_pts": np.asarray([first_downsampled_target_point_cloud],dtype=np.float32),
"combined_scanned_pts": np.asarray(first_downsampled_target_point_cloud,dtype=np.float32), "combined_scanned_pts": np.asarray(first_downsampled_target_point_cloud,dtype=np.float32),
"first_to_world_9d": np.asarray([first_to_world_9d],dtype=np.float32), "first_to_world_9d": np.asarray([first_to_world_9d],dtype=np.float32),

24
utils/reconstruction.py
View File

@ -8,9 +8,9 @@ class ReconstructionUtil:
def compute_coverage_rate(target_point_cloud, combined_point_cloud, threshold=0.01): def compute_coverage_rate(target_point_cloud, combined_point_cloud, threshold=0.01):
kdtree = cKDTree(combined_point_cloud) kdtree = cKDTree(combined_point_cloud)
distances, _ = kdtree.query(target_point_cloud) distances, _ = kdtree.query(target_point_cloud)
covered_points = np.sum(distances < threshold*2) covered_points_num = np.sum(distances < threshold)
coverage_rate = covered_points / target_point_cloud.shape[0] coverage_rate = covered_points_num / target_point_cloud.shape[0]
return coverage_rate return coverage_rate, covered_points_num
@staticmethod @staticmethod
def compute_overlap_rate(new_point_cloud, combined_point_cloud, threshold=0.01): def compute_overlap_rate(new_point_cloud, combined_point_cloud, threshold=0.01):
@ -46,10 +46,12 @@ class ReconstructionUtil:
downsampled_max_rec_pts = PtsUtil.voxel_downsample_point_cloud(max_rec_pts, threshold) downsampled_max_rec_pts = PtsUtil.voxel_downsample_point_cloud(max_rec_pts, threshold)
max_rec_pts_num = downsampled_max_rec_pts.shape[0] max_rec_pts_num = downsampled_max_rec_pts.shape[0]
max_rec_pts_coverage = ReconstructionUtil.compute_coverage_rate(target_point_cloud, downsampled_max_rec_pts, threshold) max_real_rec_pts_coverage, _ = ReconstructionUtil.compute_coverage_rate(target_point_cloud, downsampled_max_rec_pts, threshold)
new_coverage = ReconstructionUtil.compute_coverage_rate(downsampled_max_rec_pts, combined_point_cloud, threshold) new_coverage, new_covered_num = ReconstructionUtil.compute_coverage_rate(downsampled_max_rec_pts, combined_point_cloud, threshold)
current_coverage = new_coverage current_coverage = new_coverage
current_covered_num = new_covered_num
remaining_views = list(range(len(point_cloud_list))) remaining_views = list(range(len(point_cloud_list)))
view_sequence = [(init_view, current_coverage)] view_sequence = [(init_view, current_coverage)]
cnt_processed_view = 0 cnt_processed_view = 0
@ -61,6 +63,7 @@ class ReconstructionUtil:
best_view = None best_view = None
best_coverage_increase = -1 best_coverage_increase = -1
best_combined_point_cloud = None best_combined_point_cloud = None
best_covered_num = 0
for view_index in remaining_views: for view_index in remaining_views:
if point_cloud_list[view_index].shape[0] == 0: if point_cloud_list[view_index].shape[0] == 0:
@ -81,25 +84,26 @@ class ReconstructionUtil:
start = time.time() start = time.time()
new_combined_point_cloud = np.vstack([combined_point_cloud, point_cloud_list[view_index]]) new_combined_point_cloud = np.vstack([combined_point_cloud, point_cloud_list[view_index]])
new_downsampled_combined_point_cloud = PtsUtil.voxel_downsample_point_cloud(new_combined_point_cloud,threshold) new_downsampled_combined_point_cloud = PtsUtil.voxel_downsample_point_cloud(new_combined_point_cloud,threshold)
new_coverage = ReconstructionUtil.compute_coverage_rate(downsampled_max_rec_pts, new_downsampled_combined_point_cloud, threshold) new_coverage, new_covered_num = ReconstructionUtil.compute_coverage_rate(downsampled_max_rec_pts, new_downsampled_combined_point_cloud, threshold)
end = time.time() end = time.time()
#print(f"compute_coverage_rate Time: {end-start}") #print(f"compute_coverage_rate Time: {end-start}")
coverage_increase = new_coverage - current_coverage coverage_increase = new_coverage - current_coverage
if coverage_increase > best_coverage_increase: if coverage_increase > best_coverage_increase:
best_coverage_increase = coverage_increase best_coverage_increase = coverage_increase
best_view = view_index best_view = view_index
best_covered_num = new_covered_num
best_combined_point_cloud = new_downsampled_combined_point_cloud best_combined_point_cloud = new_downsampled_combined_point_cloud
if best_view is not None: if best_view is not None:
if best_coverage_increase <=1e-3: if best_coverage_increase <=1e-3 or best_covered_num - current_covered_num <= 5:
break break
selected_views.append(best_view) selected_views.append(best_view)
best_rec_pts_num = best_combined_point_cloud.shape[0] best_rec_pts_num = best_combined_point_cloud.shape[0]
print(f"Current rec pts num: {curr_rec_pts_num}, Best rec pts num: {best_rec_pts_num}, Max rec pts num: {max_rec_pts_num}") print(f"Current rec pts num: {curr_rec_pts_num}, Best rec pts num: {best_rec_pts_num}, Best cover pts: {best_covered_num}, Max rec pts num: {max_rec_pts_num}")
print(f"Current coverage: {current_coverage}, Best coverage increase: {best_coverage_increase}, Max coverage: {max_rec_pts_coverage}") print(f"Current coverage: {current_coverage}, Best coverage increase: {best_coverage_increase}, Max Real coverage: {max_real_rec_pts_coverage}")
current_covered_num = best_covered_num
curr_rec_pts_num = best_rec_pts_num curr_rec_pts_num = best_rec_pts_num
combined_point_cloud = best_combined_point_cloud combined_point_cloud = best_combined_point_cloud
remaining_views.remove(best_view) remaining_views.remove(best_view)