Functions — github.com/apple/ml-aspanformer

hub / github.com/apple/ml-aspanformer / functions

Functions198 in github.com/apple/ml-aspanformer

⨍ Functions 198◇ Types & classes 30

↓ 36 callersMethodupdate

Args: f0: [N, D, H, W] f1: [N, D, H, W] Returns: f0_new: (N, d, h, w)

src/ASpanFormer/aspan_module/transformer.py:42

↓ 10 callersFunctionconv3x3

3x3 convolution with padding

src/ASpanFormer/backbone/resnet_fpn.py:10

↓ 9 callersFunctionflattenList

(x)

src/utils/misc.py:67

↓ 7 callersFunctionconv1x1

1x1 convolution without padding

src/ASpanFormer/backbone/resnet_fpn.py:5

↓ 7 callersMethodprofile

(self, action_name: str)

src/utils/profiler.py:20

↓ 6 callersMethodload

(self, file_handle)

tools/SensorData.py:14

↓ 5 callersMethodsave_mat_to_file

(self, matrix, filename)

tools/SensorData.py:104

↓ 4 callersMethod_make_layer

(self, block, dim, stride=1)

src/ASpanFormer/backbone/resnet_fpn.py:172

↓ 4 callersMethod_setup_dataset

Setup train / val / test set

src/lightning/data.py:168

↓ 4 callersFunctionmake_matching_figures_offset

Make matching figures for a batch. Args: data (Dict): a batch updated by PL_LoFTR. config (Dict): matcher config Returns

src/utils/plotting.py:178

↓ 3 callersMethod__init__

(self,dim)

src/ASpanFormer/aspan_module/attention.py:9

↓ 3 callersMethod__init__

(self, d_model, d_flow,d_value, nhead)

src/ASpanFormer/aspan_module/transformer.py:10

↓ 3 callersMethod_make_layer

(self, block, dim, stride=1)

src/ASpanFormer/backbone/resnet_fpn.py:92

↓ 3 callersFunctionall_gather

Run all_gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object group: a torch proces

src/utils/comm.py:141

↓ 3 callersFunctiongather

Run gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object dst (int): destination ra

src/utils/comm.py:179

↓ 3 callersFunctionget_cfg_defaults

Get a yacs CfgNode object with default values for my_project.

src/config/default.py:176

↓ 3 callersFunctionget_world_size

()

src/utils/comm.py:22

↓ 3 callersFunctionload_array_from_s3

( path, client, cv_type, use_h5py=False, )

src/utils/dataset.py:18

↓ 3 callersFunctionlower_config

(yacs_cfg)

src/utils/misc.py:14

↓ 3 callersFunctionsymmetric_epipolar_distance

Squared symmetric epipolar distance. This can be seen as a biased estimation of the reprojection error. Args: pts0 (torch.Tensor): [N,

src/utils/metrics.py:30

↓ 2 callersMethod__init__

(self, in_planes, planes, stride=1)

src/ASpanFormer/backbone/resnet_fpn.py:16

↓ 2 callersFunction_build_dataset

(dataset: Dataset, *args, **kwargs)

src/lightning/data.py:325

↓ 2 callersMethod_compute_metrics

(self, batch)

src/lightning/lightning_aspanformer.py:98

↓ 2 callersFunction_compute_symmetrical_epipolar_errors_offset

Update data (dict):{"epi_errs": [M]}

src/utils/metrics.py:107

↓ 2 callersFunction_get_global_gloo_group

Return a process group based on gloo backend, containing all the ranks The result is cached.

src/utils/comm.py:84

↓ 2 callersFunction_make_evaluation_figure_offset

(data, b_id, alpha='dynamic',side='')

src/utils/plotting.py:106

↓ 2 callersFunction_pad_to_largest_tensor

Returns: list[int]: size of the tensor, on each rank Tensor: padded tensor that has the max size

src/utils/comm.py:113

↓ 2 callersMethod_read_abs_pose

(self, scene_name, name)

src/datasets/scannet.py:56

↓ 2 callersFunction_serialize_to_tensor

(data, group)

src/utils/comm.py:95

↓ 2 callersMethod_trainval_inference

(self, batch)

src/lightning/lightning_aspanformer.py:85

↓ 2 callersFunctionaggregate_metrics

Aggregate metrics for the whole dataset: (This method should be called once per dataset) 1. AUC of the pose error (angular) at the threshold

src/utils/metrics.py:235

↓ 2 callersFunctionbuild_augmentor

(method=None, **kwargs)

src/utils/augment.py:41

↓ 2 callersFunctionbuild_profiler

(name)

src/utils/profiler.py:30

↓ 2 callersFunctioncompute_symmetrical_epipolar_errors

Update: data (dict):{"epi_errs": [M]}

src/utils/metrics.py:50

↓ 2 callersFunctioncompute_symmetrical_epipolar_errors_offset_bidirectional

Update data (dict):{"epi_errs": [M]}

src/utils/metrics.py:98

↓ 2 callersMethoddecode_flow

(self,flow_feature,kshape)

src/ASpanFormer/aspan_module/transformer.py:126

↓ 2 callersFunctiondynamic_alpha

(n_matches, milestones=[0, 300, 1000, 2000], alphas=[1.0, 0.8, 0.4, 0.2])

src/utils/plotting.py:201

↓ 2 callersFunctionepidist_prec

(errors, thresholds, ret_dict=False,offset=False)

src/utils/metrics.py:221

↓ 2 callersFunctionerror_colormap

(err, thr, alpha=1.0)

src/utils/plotting.py:215

↓ 2 callersMethodflow_loss_worker

(self,flow,batch_indicies,self_indicies,cross_indicies,w)

src/losses/aspan_loss.py:30

↓ 2 callersMethodget_offset_match_work

(self, offset,conf, data,side)

src/ASpanFormer/utils/coarse_matching.py:305

↓ 2 callersFunctionget_rank

()

src/utils/comm.py:30

↓ 2 callersFunctionimread_gray

(path, augment_fn=None, client=SCANNET_CLIENT)

src/utils/dataset.py:38

↓ 2 callersMethodload_state_dict

(self, state_dict, *args, **kwargs)

src/ASpanFormer/aspanformer.py:104

↓ 2 callersFunctionmake_matching_figure

( img0, img1, mkpts0, mkpts1, color, kpts0=None, kpts1=None, text=[], dpi=75, path=None)

src/utils/plotting.py:20

↓ 2 callersFunctionmake_matching_figures

Make matching figures for a batch. Args: data (Dict): a batch updated by PL_LoFTR. config (Dict): matcher config Returns

src/utils/plotting.py:155

↓ 2 callersFunctionmask_pts_at_padded_regions

For megadepth dataset, zero-padding exists in images

src/ASpanFormer/utils/supervision.py:14

↓ 2 callersFunctionout_bound_mask

(pt, w, h)

src/ASpanFormer/utils/supervision.py:74

↓ 2 callersFunctionpad_bottom_right

(inp, pad_size, ret_mask=False)

src/utils/dataset.py:71

↓ 2 callersFunctionread_megadepth_depth

(path, pad_to=None)

src/utils/dataset.py:128

↓ 2 callersFunctionread_megadepth_gray

Args: resize (int, optional): the longer edge of resized images. None for no resize. padding (bool): If set to 'True', zero-pad r

src/utils/dataset.py:93

↓ 2 callersFunctionread_scannet_depth

(path)

src/utils/dataset.py:160

↓ 2 callersFunctionread_scannet_gray

Args: resize (tuple): align image to depthmap, in (w, h). augment_fn (callable, optional): augments images with pre-defined visua

src/utils/dataset.py:141

↓ 2 callersMethodresize_df

(self,image,df=32)

src/ASpanFormer/aspanformer.py:126

↓ 2 callersMethodupdate

(self,x0,x1,flow0,flow_feature0,pos1,mask0,mask1,ds0,ds1)

src/ASpanFormer/aspan_module/transformer.py:109

↓ 2 callersFunctionwarp_kpts

Warp kpts0 from I0 to I1 with depth, K and Rt Also check covisibility and depth consistency. Depth is consistent if relative error < 0.2 (har

src/ASpanFormer/utils/geometry.py:5

↓ 1 callersMethod__init__

(self, d_model, nhead, attention='linear')

src/ASpanFormer/aspan_module/loftr.py:7

↓ 1 callersFunction_compute_conf_thresh

(data)

src/utils/plotting.py:7

↓ 1 callersMethod_compute_fine_loss_l2

Args: expec_f (torch.Tensor): [M, 2] <x, y> expec_f_gt (torch.Tensor): [M, 2] <x, y>

src/losses/aspan_loss.py:130

↓ 1 callersMethod_compute_fine_loss_l2_std

Args: expec_f (torch.Tensor): [M, 3] <x, y, std> expec_f_gt (torch.Tensor): [M, 2] <x, y>

src/losses/aspan_loss.py:146

↓ 1 callersMethod_compute_rel_pose

(self, scene_name, name0, name1)

src/datasets/scannet.py:62

↓ 1 callersFunction_make_confidence_figure

(data, b_id)

src/utils/plotting.py:150

↓ 1 callersFunction_make_evaluation_figure

(data, b_id, alpha='dynamic')

src/utils/plotting.py:68

↓ 1 callersMethod_reset_parameters

(self)

src/ASpanFormer/aspan_module/loftr.py:81

↓ 1 callersMethod_reset_parameters

(self)

src/ASpanFormer/aspan_module/fine_preprocess.py:24

↓ 1 callersMethod_reset_parameters

(self)

src/ASpanFormer/aspan_module/transformer.py:208

↓ 1 callersFunctionbuild_backbone

(config)

src/ASpanFormer/backbone/__init__.py:4

↓ 1 callersFunctionbuild_optimizer

(model, config)

src/optimizers/__init__.py:5

↓ 1 callersFunctionbuild_scheduler

Returns: scheduler (dict):{ 'scheduler': lr_scheduler, 'interval': 'step', # or 'epoch' 'monitor': '

src/optimizers/__init__.py:17

↓ 1 callersMethodcompute_c_weight

compute element-wise weights for computing coarse-level loss.

src/losses/aspan_loss.py:177

↓ 1 callersMethodcompute_coarse_loss

Point-wise CE / Focal Loss with 0 / 1 confidence as gt. Args: conf (torch.Tensor): (N, HW0, HW1) / (N, HW0+1, HW1+1)

src/losses/aspan_loss.py:46

↓ 1 callersMethodcompute_fine_loss

(self, expec_f, expec_f_gt)

src/losses/aspan_loss.py:122

↓ 1 callersMethodcompute_flow_loss

(self,coarse_corr_gt,flow_list,h0,w0,h1,w1)

src/losses/aspan_loss.py:22

↓ 1 callersFunctioncompute_max_candidates

Compute the max candidates of all pairs within a batch Args: p_m0, p_m1 (torch.Tensor): padded masks

src/ASpanFormer/utils/coarse_matching.py:48

↓ 1 callersFunctioncompute_pose_errors

Update: data (dict):{ "R_errs" List[float]: [N] "t_errs" List[float]: [N] "inliers" List[np.ndarray]

src/utils/metrics.py:163

↓ 1 callersFunctioncompute_supervision_coarse

(data, config)

src/ASpanFormer/utils/supervision.py:112

↓ 1 callersFunctioncompute_supervision_fine

(data, config)

src/ASpanFormer/utils/supervision.py:146

↓ 1 callersMethoddecompress_color

(self, compression_type)

tools/SensorData.py:35

↓ 1 callersMethoddecompress_color_jpeg

(self)

tools/SensorData.py:42

↓ 1 callersMethoddecompress_depth

(self, compression_type)

tools/SensorData.py:24

↓ 1 callersMethoddecompress_depth_zlib

(self)

tools/SensorData.py:31

↓ 1 callersFunctionerror_auc

Args: errors (list): [N,] thresholds (list)

src/utils/metrics.py:201

↓ 1 callersFunctionestimate_pose

(kpts0, kpts1, K0, K1, thresh, conf=0.99999)

src/utils/metrics.py:134

↓ 1 callersMethodexport_color_images

(self, output_path, image_size=None, frame_skip=1)

tools/SensorData.py:93

↓ 1 callersMethodexport_depth_images

(self, output_path, image_size=None, frame_skip=1)

tools/SensorData.py:78

↓ 1 callersMethodexport_intrinsics

(self, output_path)

tools/SensorData.py:118

↓ 1 callersMethodexport_poses

(self, output_path, frame_skip=1)

tools/SensorData.py:110

↓ 1 callersMethodforward

Update: data (dict): { 'image0': (torch.Tensor): (N, 1, H, W) 'image1': (torch.Tensor): (N, 1, H

src/ASpanFormer/aspanformer.py:30

↓ 1 callersMethodget_coarse_match

Args: conf_matrix (torch.Tensor): [N, L, S] data (dict): with keys ['hw0_i', 'hw1_i', 'hw0_c', 'hw1_c'] Retur

src/ASpanFormer/utils/coarse_matching.py:158

↓ 1 callersFunctionget_divisible_wh

(w, h, df=None)

src/utils/dataset.py:63

↓ 1 callersMethodget_fine_match

(self, coords_normed, data)

src/ASpanFormer/utils/fine_matching.py:63

↓ 1 callersFunctionget_local_split

The local rank only loads a split of the dataset.

src/utils/dataloader.py:6

↓ 1 callersMethodget_offset_match

Args: offset (torch.Tensor): [L, B, H, W, 2] data (dict): with keys ['hw0_i', 'hw1_i', 'hw0_c', 'hw1_c'] Retu

src/ASpanFormer/utils/coarse_matching.py:271

↓ 1 callersFunctionget_rank_zero_only_logger

(logger: _Logger)

src/utils/misc.py:32

↓ 1 callersFunctionget_resized_wh

(w, h, resize=None)

src/utils/dataset.py:54

↓ 1 callersMethodgroup_attention

(self,query,key,value,temp,mask_sample=None)

src/ASpanFormer/aspan_module/attention.py:118

↓ 1 callersMethodload

(self, filename)

tools/SensorData.py:53

↓ 1 callersFunctionlower_config

(yacs_cfg)

src/ASpanFormer/utils/cvpr_ds_config.py:4

↓ 1 callersFunctionmain

()

train.py:66

↓ 1 callersFunctionmain

()

tools/reader.py:21

next →1–100 of 198, ranked by callers