Functions — github.com/ByteDance-Seed/Depth-Anything-3

hub / github.com/ByteDance-Seed/Depth-Anything-3 / functions

Functions731 in github.com/ByteDance-Seed/Depth-Anything-3

⨍ Functions 731◇ Types & classes 70 ↳ Endpoints 14

The following function was copied from: https://github.com/MichaelGrupp/evo/blob/3067541b350528fe46375423e5bc3a7c42c06c63/evo/core/geometry.p

da3_streaming/loop_utils/loop_refinement.py:83

↓ 2 callersMethodupdate_depth_view

Update depth view for a specific view index. Args: processed_data: Processed data dictionary view_index: Ind

src/depth_anything_3/app/modules/visualization.py:81

↓ 2 callersMethodupdate_measure_view

Update measure view for a specific view index. Args: processed_data: Processed data dictionary view_index: I

src/depth_anything_3/app/modules/event_handlers.py:506

↓ 2 callersFunctionweighted_estimate_se3

source_points: (Nx3) target_points: (Nx3) :weights: (N,) [0,1]

da3_streaming/loop_utils/sim3utils.py:606

↓ 2 callersFunctionweighted_estimate_se3_torch

(source_points, target_points, weights)

da3_streaming/loop_utils/alignment_torch.py:21

↓ 2 callersFunctionweighted_estimate_sim3

source_points: (Nx3) target_points: (Nx3) :weights: (N,) [0,1]

da3_streaming/loop_utils/sim3utils.py:641

↓ 2 callersFunctionweighted_estimate_sim3_numba

(source_points, target_points, weights, align_method="sim3")

da3_streaming/loop_utils/sim3utils.py:805

↓ 2 callersFunctionweighted_estimate_sim3_numba_torch

(source_points, target_points, weights, align_method="sim3")

da3_streaming/loop_utils/alignment_torch.py:86

↓ 2 callersFunctionweighted_estimate_sim3_numba_triton

( source_points, target_points, weights, align_method="sim3" )

da3_streaming/loop_utils/alignment_triton.py:354

↓ 2 callersFunctionweighted_estimate_sim3_torch

(source_points, target_points, weights)

da3_streaming/loop_utils/alignment_torch.py:51

↓ 2 callersFunctionwrite_ply_batch

(f, points, colors)

da3_streaming/loop_utils/sim3utils.py:431

↓ 2 callersFunctionwrite_ply_header

(f, num_vertices)

da3_streaming/loop_utils/sim3utils.py:415

↓ 1 callersMethod__init__

Initialize DepthAnything3Net with given yaml-initialized configuration.

src/depth_anything_3/model/da3.py:66

↓ 1 callersMethod__init__

Initializes the position generator with an empty cache.

src/depth_anything_3/model/dinov2/layers/rope.py:34

↓ 1 callersMethod__init__

( self, in_features: int, hidden_features: Optional[int] = None, out_features:

src/depth_anything_3/model/dinov2/layers/swiglu_ffn.py:13

↓ 1 callersFunction_add_cameras_to_scene

Draws camera frustums to visualize their position and orientation. This function renders each camera as a wireframe pyramid, originating from

src/depth_anything_3/utils/export/glb.py:322

↓ 1 callersMethod_add_processed_images

Add processed images to prediction for visualization.

src/depth_anything_3/api.py:399

↓ 1 callersMethod_align_to_input_extrinsics_intrinsics

Align depth map to input extrinsics

src/depth_anything_3/api.py:341

↓ 1 callersMethod_apply_depth_alignment

Apply depth alignment using least squares scaling.

src/depth_anything_3/model/da3.py:385

↓ 1 callersMethod_apply_metric_scaling

Apply metric scaling to the metric depth output.

src/depth_anything_3/model/da3.py:374

↓ 1 callersMethod_apply_nms_filter

Apply Non-Maximum Suppression (NMS) filtering to loop pairs

da3_streaming/loop_utils/loop_detector.py:208

↓ 1 callersMethod_apply_sky_activation

Sky head activation (fixed 1 channel): * 'sigmoid' -> Sigmoid probability map * 'relu' -> ReLU positive domain output

src/depth_anything_3/model/dpt.py:311

↓ 1 callersMethod_build_proj_mats

Compute per-view 4x4 projection matrices from K and [R|t].

src/depth_anything_3/bench/datasets/dtu.py:626

↓ 1 callersFunction_camera_frustum_lines

( K: np.ndarray, ext_w2c: np.ndarray, W: int, H: int, scale: float )

src/depth_anything_3/utils/export/glb.py:362

↓ 1 callersMethod_cap_points

Downsample points if exceeding max count.

src/depth_anything_3/bench/datasets/dtu.py:673

↓ 1 callersMethod_check_backend_status

Check backend status

src/depth_anything_3/services/inference_service.py:170

↓ 1 callersFunction_compute_alignment_transform_first_cam_glTF_center_by_points

Computes the transformation matrix to align the scene with glTF standards. This function calculates a 4x4 homogeneous matrix that centers the sce

src/depth_anything_3/utils/export/glb.py:275

↓ 1 callersMethod_compute_frequency_components

Computes frequency components for rotary embeddings. Args: dim: Feature dimension (must be even). seq_len: Maximum se

src/depth_anything_3/model/dinov2/layers/rope.py:87

↓ 1 callersMethod_compute_pose_with_gt

Compute pose metrics using saved GT meta (handles frame sampling). Args: result_path: Path to npz with predicted extrins

src/depth_anything_3/bench/evaluator.py:405

↓ 1 callersMethod_convert_to_prediction

Convert raw model output to Prediction object.

src/depth_anything_3/api.py:391

↓ 1 callersFunction_create_xyf

Creates a grid of pixel coordinates and frame indices (fidx) for all frames.

src/depth_anything_3/utils/export/colmap.py:130

↓ 1 callersMethod_crop_ixt

( self, intrinsic: np.ndarray | None, orig_w: int, orig_h: int, w: int

src/depth_anything_3/utils/io/input_processor.py:278

↓ 1 callersMethod_depth_mask_path

Get path to depth mask for a scene and frame.

src/depth_anything_3/bench/datasets/dtu.py:553

↓ 1 callersMethod_ensure_decending_order

(self, tuples_list)

da3_streaming/loop_utils/loop_detector.py:239

↓ 1 callersFunction_estimate_scene_scale

(points: np.ndarray, fallback: float = 1.0)

src/depth_anything_3/utils/export/glb.py:266

↓ 1 callersMethod_eval_pose

Compute pose-estimation metrics for each dataset and scene.

src/depth_anything_3/bench/evaluator.py:260

↓ 1 callersMethod_evaluate_reconstruction

Compute accuracy, completeness, and overall metrics for one scan. Args: scanid: Scan identifier pred_ply: Pr

src/depth_anything_3/bench/datasets/dtu.py:380

↓ 1 callersMethod_export_results

Export results to specified format and directory.

src/depth_anything_3/api.py:414

↓ 1 callersMethod_extract_aux

Extract auxiliary data from model output and convert to numpy. Args: model_output: Model output dictionary Retu

src/depth_anything_3/utils/io/output_processor.py:150

↓ 1 callersMethod_extract_auxiliary_features

Extract auxiliary features from specified layers.

src/depth_anything_3/model/da3.py:286

↓ 1 callersMethod_extract_conf

Extract confidence tensor from model output and convert to numpy. Args: model_output: Model output dictionary R

src/depth_anything_3/utils/io/output_processor.py:90

↓ 1 callersMethod_extract_depth

Extract depth tensor from model output and convert to numpy. Args: model_output: Model output dictionary Return

src/depth_anything_3/utils/io/output_processor.py:77

↓ 1 callersMethod_extract_extrinsics

Extract extrinsics tensor from model output and convert to numpy. Args: model_output: Model output dictionary R

src/depth_anything_3/utils/io/output_processor.py:105

↓ 1 callersMethod_extract_intrinsics

Extract intrinsics tensor from model output and convert to numpy. Args: model_output: Model output dictionary R

src/depth_anything_3/utils/io/output_processor.py:120

↓ 1 callersMethod_extract_points

Extract masked points from a dense grid.

src/depth_anything_3/bench/datasets/dtu.py:362

↓ 1 callersMethod_extract_sky

Extract sky tensor from model output and convert to numpy. Args: model_output: Model output dictionary Returns:

src/depth_anything_3/utils/io/output_processor.py:135

↓ 1 callersFunction_filter_and_downsample

(points: np.ndarray, colors: np.ndarray, num_max: int)

src/depth_anything_3/utils/export/glb.py:255

↓ 1 callersMethod_filter_depth

Compute geometric consistency between reference and source depths. Args: ref_depth: Reference depth [1, 1, H, W]

src/depth_anything_3/bench/datasets/dtu.py:334

↓ 1 callersMethod_fuse

Feature pyramid fusion. Returns: fused_main: Tensor at finest scale (after refinenet1) aux_pyr: List of au

src/depth_anything_3/model/dualdpt.py:270

↓ 1 callersMethod_fuse_consistent_points

Fuse points consistent across multiple source views for a reference index.

src/depth_anything_3/bench/datasets/dtu.py:637

↓ 1 callersMethod_get_intermediate_layers_not_chunked

(self, x, n=1, export_feat_layers=[], **kwargs)

src/depth_anything_3/model/dinov2/vision_transformer.py:300

↓ 1 callersMethod_get_model_device

Get the device where the model is located. Returns: Device where the model parameters are located Raises:

src/depth_anything_3/api.py:423

↓ 1 callersMethod_group_by_dataset

Group metrics by dataset.

src/depth_anything_3/bench/print_metrics.py:256

↓ 1 callersMethod_handle_sky_regions

Handle sky regions by setting them to maximum depth.

src/depth_anything_3/model/da3.py:418

↓ 1 callersMethod_homo_warping

Homography warping for multi-view consistency checking. Args: src_fea: Source features [B, C, H, W] src_proj

src/depth_anything_3/bench/datasets/dtu.py:278

↓ 1 callersFunction_hsv_to_rgb

(h: float, s: float, v: float)

src/depth_anything_3/utils/export/glb.py:413

↓ 1 callersFunction_index_color_rgb

(i: int, n: int)

src/depth_anything_3/utils/export/glb.py:406

↓ 1 callersMethod_input_transform

Create image transformation function

da3_streaming/loop_utils/loop_detector.py:108

↓ 1 callersFunction_intersperse

(iterable: Iterable, delimiter: Any)

src/depth_anything_3/utils/layout_helpers.py:55

↓ 1 callersFunction_is_plain_name

(name: str)

src/depth_anything_3/services/gallery.py:637

↓ 1 callersFunction_load_gallery_html

Load and modify gallery HTML to work under /gallery/ subdirectory. Replaces API paths from root to /gallery/ prefix.

src/depth_anything_3/services/backend.py:458

↓ 1 callersMethod_load_gt_meta

Load saved GT extrinsics/intrinsics for evaluation. Returns: Dict with extrinsics and intrinsics, or None if not found

src/depth_anything_3/bench/evaluator.py:389

↓ 1 callersMethod_load_gt_meta

Load saved GT meta for fusion.

src/depth_anything_3/bench/datasets/scannetpp.py:331

↓ 1 callersMethod_load_gt_meta

Load saved GT meta (extrinsics, intrinsics, image_files) for fusion. This is needed when frames are sampled, so fuse3d uses the corr

src/depth_anything_3/bench/datasets/sevenscenes.py:207

↓ 1 callersMethod_load_gt_meta

Load saved GT meta for fusion.

src/depth_anything_3/bench/datasets/hiroom.py:209

↓ 1 callersMethod_load_gt_meta

Load saved GT meta (extrinsics, intrinsics, image_files) for fusion. This is needed when frames are sampled, so fuse3d uses the corr

src/depth_anything_3/bench/datasets/eth3d.py:301

↓ 1 callersMethod_load_image

(self, img: np.ndarray | Image.Image | str)

src/depth_anything_3/utils/io/input_processor.py:298

↓ 1 callersMethod_load_metrics

Load evaluation metrics from JSON files.

src/depth_anything_3/bench/evaluator.py:510

↓ 1 callersFunction_load_scene_list

Load scene list from file.

src/depth_anything_3/bench/datasets/hiroom.py:56

↓ 1 callersMethod_make_aux_out1_block

Factory for the aux pre-head stack before the final 1x1 projection.

src/depth_anything_3/model/dualdpt.py:321

↓ 1 callersMethod_make_divisible_by_crop

Floor each dimension to the nearest multiple of PATCH_SIZE via center crop. Example: 504x377 -> 504x364

src/depth_anything_3/utils/io/input_processor.py:351

↓ 1 callersMethod_make_divisible_by_resize

Round each dimension to nearest multiple of PATCH_SIZE via small resize.

src/depth_anything_3/utils/io/input_processor.py:365

↓ 1 callersFunction_median_nn_thresh

(pose_ref, pose_est_aligned)

src/depth_anything_3/utils/pose_align.py:101

↓ 1 callersMethod_normalize_extrinsics

Normalize extrinsics

src/depth_anything_3/api.py:327

↓ 1 callersMethod_normalize_image

(self, img: Image.Image)

src/depth_anything_3/utils/io/input_processor.py:309

↓ 1 callersMethod_normalize_rgb

(self, PCs_raw: torch.Tensor)

src/depth_anything_3/utils/pca_utils.py:233

↓ 1 callersMethod_parse_cameras_txt

Parse COLMAP-style cameras.txt file. Returns: Dict mapping camera_id to intrinsic parameters

src/depth_anything_3/bench/datasets/eth3d.py:104

↓ 1 callersMethod_parse_images_txt

Parse COLMAP-style images.txt file. Returns: Dict mapping image path to pose parameters

src/depth_anything_3/bench/datasets/eth3d.py:133

↓ 1 callersFunction_poses_from_ext

(ext_ref, ext_est)

src/depth_anything_3/utils/pose_align.py:76

↓ 1 callersMethod_prep_posed

Prepare depths/intrinsics/extrinsics for recon_posed mode.

src/depth_anything_3/bench/datasets/scannetpp.py:490

↓ 1 callersMethod_prep_posed

Prepare depths/intrinsics/extrinsics for recon_posed mode. Uses GT intrinsics/extrinsics but aligns scale via Umeyama. Same

src/depth_anything_3/bench/datasets/dtu.py:593

↓ 1 callersMethod_prep_posed

Prepare depths/intrinsics/extrinsics for recon_posed mode. Uses GT intrinsics/extrinsics but aligns depth scale via Umeyama.

src/depth_anything_3/bench/datasets/sevenscenes.py:353

↓ 1 callersMethod_prep_posed

Prepare depths/intrinsics/extrinsics for recon_posed mode.

src/depth_anything_3/bench/datasets/hiroom.py:350

↓ 1 callersMethod_prep_posed

Prepare depths/intrinsics/extrinsics for recon_posed mode. Uses GT intrinsics/extrinsics but aligns depth scale via Umeyama.

src/depth_anything_3/bench/datasets/eth3d.py:459

↓ 1 callersMethod_prep_unposed

Prepare depths/intrinsics/extrinsics for recon_unposed mode.

src/depth_anything_3/bench/datasets/scannetpp.py:431

↓ 1 callersMethod_prep_unposed

Prepare depths/intrinsics/extrinsics for recon_unposed mode. Applies Umeyama scale, rescales intrinsics if depth resolution

src/depth_anything_3/bench/datasets/dtu.py:559

↓ 1 callersMethod_prep_unposed

Prepare depths/intrinsics/extrinsics for recon_unposed mode. Similar to ETH3D but uses GT depth for masking instead of separate mask

src/depth_anything_3/bench/datasets/sevenscenes.py:301

↓ 1 callersMethod_prep_unposed

Prepare depths/intrinsics/extrinsics for recon_unposed mode.

src/depth_anything_3/bench/datasets/hiroom.py:297

↓ 1 callersMethod_prep_unposed

Prepare depths/intrinsics/extrinsics for recon_unposed mode. Pipeline: 1. Umeyama scale alignment 2. Load GT mask fo

src/depth_anything_3/bench/datasets/eth3d.py:396

↓ 1 callersMethod_prepare_model_inputs

Prepare tensors for model input.

src/depth_anything_3/api.py:301

↓ 1 callersMethod_prepare_rope

(self, B, S, H, W, device)

src/depth_anything_3/model/dinov2/vision_transformer.py:282

↓ 1 callersMethod_preprocess_inputs

Preprocess input images using input processor.

src/depth_anything_3/api.py:275

↓ 1 callersMethod_print_dataset_section

Print metrics section for a single dataset.

src/depth_anything_3/bench/print_metrics.py:271

↓ 1 callersMethod_print_summary

Print summary table with key metrics across all datasets. Format: One row per metric, datasets as columns. Order: HiRoom, ET

src/depth_anything_3/bench/print_metrics.py:347

↓ 1 callersMethod_process_camera_estimation

Process camera pose estimation if camera decoder is available.

src/depth_anything_3/model/da3.py:211

↓ 1 callersMethod_process_depth_head

Process features through the depth prediction head.

src/depth_anything_3/model/da3.py:205

↓ 1 callersMethod_process_gs_head

Process 3DGS parameters estimation if 3DGS head is available.

src/depth_anything_3/model/da3.py:230

↓ 1 callersMethod_process_images

Process uploaded images. Args: input_images: List of input image files target_dir_images: Target directory f

src/depth_anything_3/app/modules/file_handlers.py:101

↓ 1 callersMethod_process_mono_sky_estimation

Process mono sky estimation.

src/depth_anything_3/model/da3.py:155

↓ 1 callersMethod_process_ray_pose_estimation

Process ray pose estimation if ray pose decoder is available.

src/depth_anything_3/model/da3.py:181

← previous next →201–300 of 731, ranked by callers