MCPcopy Create free account

hub / github.com/BIT-MJY/SeqOT / functions

Functions49 in github.com/BIT-MJY/SeqOT

↓ 6 callersFunctionread_one_need_descriptor_from_seq_ft
(file_num, descriptors, seq_len, poses=None)
tools/read_samples.py:96
↓ 5 callersFunctionrange_projection
Project a pointcloud into a spherical projection, range image. Args: current_vertex: raw point clouds Returns: proj_range: pr
tools/utils/utils.py:60
↓ 4 callersFunctionbest_pos_distance
(query, pos_vecs)
tools/loss.py:7
↓ 4 callersFunctionget_velo
(velofile)
tools/utils/com_overlap_yaw.py:33
↓ 4 callersFunctionload_files
Load all files in a folder and sort.
tools/utils/utils.py:240
↓ 3 callersFunctionread_one_need_from_seq
(file_num, seq_len, poses=None, range_image_root=None)
tools/read_samples.py:17
↓ 3 callersMethodtrain
(self)
train/training_gem.py:56
↓ 2 callersFunctioneuler_angles_from_rotation_matrix
From the paper by Gregory G. Slabaugh, Computing Euler angles from a rotation matrix psi, theta, phi = roll pitch yaw (x, y, z) Arg
tools/utils/utils.py:196
↓ 2 callersFunctionisclose
(x, y, rtol=1.e-5, atol=1.e-8)
tools/utils/utils.py:206
↓ 1 callersMethod__init__
(self, feature_size, max_samples, cluster_size, output_dim, gating=True, add_batch_norm=True,
modules/netvlad.py:8
↓ 1 callersFunctioncal_topN
(prediction_file_name, ground_truth_file_name, topn)
test/test_seqot.py:19
↓ 1 callersFunctioncom_overlap_yaw
compute the overlap and yaw ground truth from the ground truth poses, which is used for OverlapNet training and testing. Args: scan_p
tools/utils/com_overlap_yaw.py:38
↓ 1 callersFunctiondata2xyzi
(data, flip=True)
tools/utils/com_overlap_yaw.py:19
↓ 1 callersFunctiondata2xyzi
(data, flip=True)
tools/utils/gen_depth_data.py:24
↓ 1 callersMethodeval
(self)
test/test_gem_prepare.py:38
↓ 1 callersMethodeval
(self)
visualize/viz.py:43
↓ 1 callersMethodeval
(self)
train/gen_sub_descriptors.py:40
↓ 1 callersMethodgem
(self, x, p=3, eps=1e-6)
modules/gem.py:15
↓ 1 callersFunctiongen_depth_data
Generate projected range data in the shape of (64, 900, 1). The input raw data are in the shape of (Num_points, 3).
tools/utils/gen_depth_data.py:42
↓ 1 callersFunctionget_velo
(velofile)
tools/utils/gen_depth_data.py:38
↓ 1 callersFunctionmain
(topn, ground_truth_file_name)
test/test_seqot.py:52
↓ 1 callersFunctionnormalize_data
Normalize the training data according to the overlap value. Args: ground_truth_mapping: the raw ground truth mapping array Returns:
tools/utils/normalize_data.py:9
↓ 1 callersFunctionread_one_batch_pos_neg
(f1_index, f1_seq, train_imgf1, train_imgf2, train_dir1, train_dir2, range_image_root,
tools/read_samples.py:43
↓ 1 callersFunctionread_one_batch_pos_neg_descriptors
(f1_index, f1_seq, train_imgf1, train_imgf2, train_dir1, train_dir2, train_overlap, overlap_thresh, seq_len, d
tools/read_samples.py:112
↓ 1 callersFunctionsplit_train_val
Split the ground truth data into training and validation two parts. Args: ground_truth_mapping: the raw ground truth mapping array Re
tools/utils/split_train_val.py:10
↓ 1 callersMethodtrain
(self)
train/training_seqot.py:53
↓ 1 callersFunctiontriplet_loss
(q_vec, pos_vecs, neg_vecs, margin, use_min=False, lazy=False, ignore_zero_loss=False)
tools/loss.py:39
↓ 1 callersFunctionwrap
Wrap the boarder of the range image.
tools/utils/utils.py:185
Method__init__
(self, seqL=5)
modules/seqTransformerCat.py:17
Method__init__
(self, p=3, eps=1e-6)
modules/gem.py:7
Method__init__
(self, dim, add_batch_norm=True)
modules/netvlad.py:76
Method__init__
(self, seqlen=20, pretrained_weights=None, descs_database=None, descs_query=None)
test/test_gem_prepare.py:27
Method__init__
(self, seqlen=20, pretrained_weights=None, descs_database=None, descs_query=None, gt_file_nam
visualize/viz.py:28
Method__init__
(self, height=32, width=900, seqlen=3, lr=0.000005, resume=False, pretrained_weights=None, tr
train/training_seqot.py:28
Method__init__
(self, seqlen=3, pretrained_weights=None,range_image_database_root=None, range_
train/gen_sub_descriptors.py:26
Method__init__
(self, lr=0.000005, seqlen=20, resume=False, pretrained_weights=None, descriptors=None, train_set=None, poses_
train/training_gem.py:28
Method__repr__
(self)
modules/gem.py:19
Functioncom_overlap_yaw_idx
(scan_paths, poses, frame_idx, leg_output_width=360, viz=False, idx_in_range=None)
tools/utils/com_overlap_yaw.py:128
Methodforward
(self, x_l)
modules/seqTransformerCat.py:50
Methodforward
(self, x)
modules/gem.py:12
Methodforward
(self, x)
modules/netvlad.py:41
Methodforward
(self, x)
modules/netvlad.py:92
Functiongen_normal_map
Generate a normal image given the range projection of a point cloud. Args: current_range: range projection of a point cloud, each pixe
tools/utils/utils.py:140
Functionload_calib
Load calibrations (T_cam_velo) from file.
tools/utils/utils.py:39
Functionload_poses
Load ground truth poses (T_w_cam0) from file. Args: pose_path: (Complete) filename for the pose file Returns: A numpy a
tools/utils/utils.py:11
Functionload_vertex
Load 3D points of a scan. The fileformat is the .bin format used in the KITTI dataset. Args: scan_path: the (full) filename of t
tools/utils/utils.py:224
Functionsliding_triplet_loss
(q_vec, pos_vecs, neg_vecs, margin, use_min=True, lazy=False, ignore_zero_loss=False)
tools/loss.py:17
Functiontriplet_loss_inv
(q_vec, pos_vecs, neg_vecs, margin, use_min=True, lazy=False, ignore_zero_loss=False)
tools/loss.py:70
Functiontriplet_loss_wrapper
(q_vec, pos_vecs, neg_vecs, m1, m2, use_min=False, lazy=False, ignore_zero_loss=False)
tools/loss.py:99