Adapt to the new interface for pycolmap 3.12 (#467)

* Adapt to the new pycolmap inteface

* update

* inliers to inlier_mask

* specify minimum pycolmap version

* fix long-term broken interface

* reconstruction.reg_image_ids() becomes set rather than list

* adapt to the new pycolmap localization interface

* format

* Fix naming error: image_list -> image_names

* Fix 3D viz and demo

* Improve logging --> progressbar for reconstuction

* Fix inloc pose estimation

* Add frames and rigs to shutil.move after reconstruction

* Fix formatting

* Fix geometric verification in triangulation

* Require pycolmap version 3.12

* Fix visualization

* Reduce size of demo notebook and switch to ALIKED as default

* Create cameras using kwargs instead of dicts

* Add write_poses to utils/io and use it in localize_sfm and localize_inloc

* Wrap incremental_mapping in OutputCapture

* Remove unused verbose argument

* tag pycolmap >=3.12.3.

---------

Co-authored-by: Phil26AT <phil.lindenberger@gmail.com>
Co-authored-by: Paul-Edouard Sarlin <15985472+sarlinpe@users.noreply.github.com>

Author: 3 people

demo.ipynb

@@ -21,7 +21,7 @@
 except ImportError:
     logger.warning("pycolmap is not installed, some features may not work.")
 else:
-    min_version = version.parse("0.6.0")
+    min_version = version.parse("3.12.3")
     found_version = pycolmap.__version__
     if found_version != "dev":
         version = version.parse(found_version)

hloc/__init__.py

@@ -11,6 +11,7 @@
 from tqdm import tqdm
 
 from . import logger
+from .utils.io import write_poses
 from .utils.parsers import names_to_pair, parse_retrieval
 
 
@@ -110,7 +111,11 @@ def pose_from_cluster(dataset_dir, q, retrieved, feature_file, match_file, skip=
         "height": height,
         "params": [focal_length, cx, cy],
     }
-    ret = pycolmap.absolute_pose_estimation(all_mkpq, all_mkp3d, cfg, 48.00)
+    estimation_options = pycolmap.AbsolutePoseEstimationOptions()
+    estimation_options.ransac.max_error = 48
+    ret = pycolmap.estimate_and_refine_absolute_pose(
+        all_mkpq, all_mkp3d, cfg, estimation_options
+    )
     ret["cfg"] = cfg
     return ret, all_mkpq, all_mkpr, all_mkp3d, all_indices, num_matches
 
@@ -140,7 +145,7 @@ def main(dataset_dir, retrieval, features, matches, results, skip_matches=None):
             dataset_dir, q, db, feature_file, match_file, skip_matches
         )
 
-        poses[q] = (ret["qvec"], ret["tvec"])
+        poses[q] = ret["cam_from_world"]
         logs["loc"][q] = {
             "db": db,
             "PnP_ret": ret,
@@ -152,13 +157,7 @@ def main(dataset_dir, retrieval, features, matches, results, skip_matches=None):
         }
 
     logger.info(f"Writing poses to {results}...")
-    with open(results, "w") as f:
-        for q in queries:
-            qvec, tvec = poses[q]
-            qvec = " ".join(map(str, qvec))
-            tvec = " ".join(map(str, tvec))
-            name = q.split("/")[-1]
-            f.write(f"{name} {qvec} {tvec}\n")
+    write_poses(poses, results, prepend_camera_name=False)
 
     logs_path = f"{results}_logs.pkl"
     logger.info(f"Writing logs to {logs_path}...")

hloc/localize_inloc.py

@@ -9,7 +9,7 @@
 from tqdm import tqdm
 
 from . import logger
-from .utils.io import get_keypoints, get_matches
+from .utils.io import get_keypoints, get_matches, write_poses
 from .utils.parsers import parse_image_lists, parse_retrieval
 
 
@@ -58,7 +58,7 @@ def __init__(self, reconstruction, config=None):
     def localize(self, points2D_all, points2D_idxs, points3D_id, query_camera):
         points2D = points2D_all[points2D_idxs]
         points3D = [self.reconstruction.points3D[j].xyz for j in points3D_id]
-        ret = pycolmap.absolute_pose_estimation(
+        ret = pycolmap.estimate_and_refine_absolute_pose(
             points2D,
             points3D,
             query_camera,
@@ -208,14 +208,7 @@ def main(
 
     logger.info(f"Localized {len(cam_from_world)} / {len(queries)} images.")
     logger.info(f"Writing poses to {results}...")
-    with open(results, "w") as f:
-        for query, t in cam_from_world.items():
-            qvec = " ".join(map(str, t.rotation.quat[[3, 0, 1, 2]]))
-            tvec = " ".join(map(str, t.translation))
-            name = query.split("/")[-1]
-            if prepend_camera_name:
-                name = query.split("/")[-2] + "/" + name
-            f.write(f"{name} {qvec} {tvec}\n")
+    write_poses(cam_from_world, results, prepend_camera_name=prepend_camera_name)
 
     logs_path = f"{results}_logs.pkl"
     logger.info(f"Writing logs to {logs_path}...")

hloc/localize_sfm.py

@@ -11,7 +11,14 @@
 
 def scene_coordinates(p2D, R_w2c, t_w2c, depth, camera):
     assert len(depth) == len(p2D)
-    p2D_norm = np.stack(pycolmap.Camera(camera._asdict()).image_to_world(p2D))
+    pycolmap_camera = pycolmap.Camera(
+        camera_id=camera.id,
+        model=camera.model,
+        width=camera.width,
+        height=camera.height,
+        params=camera.params,
+    )
+    p2D_norm = pycolmap_camera.cam_from_img(p2D)
     p2D_h = np.concatenate([p2D_norm, np.ones_like(p2D_norm[:, :1])], 1)
     p3D_c = p2D_h * depth[:, None]
     p3D_w = (p3D_c - t_w2c) @ R_w2c
@@ -52,7 +59,14 @@ def project_to_image(p3D, R, t, camera, eps: float = 1e-4, pad: int = 1):
     p3D = (p3D @ R.T) + t
     visible = p3D[:, -1] >= eps  # keep points in front of the camera
     p2D_norm = p3D[:, :-1] / p3D[:, -1:].clip(min=eps)
-    p2D = np.stack(pycolmap.Camera(camera._asdict()).world_to_image(p2D_norm))
+    pycolmap_camera = pycolmap.Camera(
+        camera_id=camera.id,
+        model=camera.model,
+        width=camera.width,
+        height=camera.height,
+        params=camera.params,
+    )
+    p2D = pycolmap_camera.img_from_cam(p2D_norm)
     size = np.array([camera.width - pad - 1, camera.height - pad - 1])
     valid = np.all((p2D >= pad) & (p2D <= size), -1)
     valid &= visible

hloc/pipelines/7Scenes/create_gt_sfm.py

@@ -5,6 +5,7 @@
 from typing import Any, Dict, List, Optional
 
 import pycolmap
+import tqdm
 
 from . import logger
 from .triangulation import (
@@ -46,7 +47,7 @@ def import_images(
             database_path,
             image_dir,
             camera_mode,
-            image_list=image_list or [],
+            image_names=image_list or [],
             options=options,
         )
 
@@ -60,6 +61,40 @@ def get_image_ids(database_path: Path) -> Dict[str, int]:
     return images
 
 
+def incremental_mapping(
+    database_path: Path,
+    image_dir: Path,
+    sfm_path: Path,
+    options: Optional[Dict[str, Any]] = None,
+) -> dict[int, pycolmap.Reconstruction]:
+    num_images = pycolmap.Database(database_path).num_images
+    pbars = []
+
+    def restart_progress_bar():
+        if len(pbars) > 0:
+            pbars[-1].close()
+        pbars.append(
+            tqdm.tqdm(
+                total=num_images,
+                desc=f"Reconstruction {len(pbars)}",
+                unit="images",
+                postfix="registered",
+            )
+        )
+        pbars[-1].update(2)
+
+    reconstructions = pycolmap.incremental_mapping(
+        database_path,
+        image_dir,
+        sfm_path,
+        options=options or {},
+        initial_image_pair_callback=restart_progress_bar,
+        next_image_callback=lambda: pbars[-1].update(1),
+    )
+
+    return reconstructions
+
+
 def run_reconstruction(
     sfm_dir: Path,
     database_path: Path,
@@ -73,11 +108,11 @@ def run_reconstruction(
     if options is None:
         options = {}
     options = {"num_threads": min(multiprocessing.cpu_count(), 16), **options}
+
     with OutputCapture(verbose):
-        with pycolmap.ostream():
-            reconstructions = pycolmap.incremental_mapping(
-                database_path, image_dir, models_path, options=options
-            )
+        reconstructions = incremental_mapping(
+            database_path, image_dir, models_path, options=options
+        )
 
     if len(reconstructions) == 0:
         logger.error("Could not reconstruct any model!")
@@ -96,7 +131,13 @@ def run_reconstruction(
         f"Largest model is #{largest_index} " f"with {largest_num_images} images."
     )
 
-    for filename in ["images.bin", "cameras.bin", "points3D.bin"]:
+    for filename in [
+        "images.bin",
+        "cameras.bin",
+        "points3D.bin",
+        "frames.bin",
+        "rigs.bin",
+    ]:
         if (sfm_dir / filename).exists():
             (sfm_dir / filename).unlink()
         shutil.move(str(models_path / str(largest_index) / filename), str(sfm_dir))
@@ -124,6 +165,9 @@ def main(
     sfm_dir.mkdir(parents=True, exist_ok=True)
     database = sfm_dir / "database.db"
 
+    logger.info(f"Writing COLMAP logs to {sfm_dir / 'colmap.LOG.*'}")
+    pycolmap.logging.set_log_destination(pycolmap.logging.INFO, sfm_dir / "colmap.LOG.")
+
     create_empty_db(database)
     import_images(image_dir, database, camera_mode, image_list, image_options)
     image_ids = get_image_ids(database)

hloc/reconstruction.py

@@ -1,7 +1,4 @@
 import argparse
-import contextlib
-import io
-import sys
 from pathlib import Path
 from typing import Any, Dict, List, Optional
 
@@ -22,15 +19,11 @@ def __init__(self, verbose: bool):
 
     def __enter__(self):
         if not self.verbose:
-            self.capture = contextlib.redirect_stdout(io.StringIO())
-            self.out = self.capture.__enter__()
+            pycolmap.logging.alsologtostderr = False
 
     def __exit__(self, exc_type, *args):
         if not self.verbose:
-            self.capture.__exit__(exc_type, *args)
-            if exc_type is not None:
-                logger.error("Failed with output:\n%s", self.out.getvalue())
-        sys.stdout.flush()
+            pycolmap.logging.alsologtostderr = True
 
 
 def create_db_from_model(
@@ -114,12 +107,11 @@ def estimation_and_geometric_verification(
 ):
     logger.info("Performing geometric verification of the matches...")
     with OutputCapture(verbose):
-        with pycolmap.ostream():
-            pycolmap.verify_matches(
-                database_path,
-                pairs_path,
-                options=dict(ransac=dict(max_num_trials=20000, min_inlier_ratio=0.1)),
-            )
+        pycolmap.verify_matches(
+            database_path,
+            pairs_path,
+            options=dict(ransac=dict(max_num_trials=20000, min_inlier_ratio=0.1)),
+        )
 
 
 def geometric_verification(
@@ -170,7 +162,7 @@ def geometric_verification(
                 db.add_two_view_geometry(id0, id1, matches)
                 continue
 
-            cam1_from_cam0 = image1.cam_from_world * image0.cam_from_world.inverse()
+            cam1_from_cam0 = image1.cam_from_world() * image0.cam_from_world().inverse()
             errors0, errors1 = compute_epipolar_errors(
                 cam1_from_cam0, kps0[matches[:, 0]], kps1[matches[:, 1]]
             )
@@ -207,10 +199,9 @@ def run_triangulation(
     if options is None:
         options = {}
     with OutputCapture(verbose):
-        with pycolmap.ostream():
-            reconstruction = pycolmap.triangulate_points(
-                reference_model, database_path, image_dir, model_path, options=options
-            )
+        reconstruction = pycolmap.triangulate_points(
+            reference_model, database_path, image_dir, model_path, options=options
+        )
     return reconstruction
 
 

hloc/triangulation.py

@@ -7,7 +7,7 @@ def to_homogeneous(p):
 
 
 def compute_epipolar_errors(j_from_i: pycolmap.Rigid3d, p2d_i, p2d_j):
-    j_E_i = j_from_i.essential_matrix()
+    j_E_i = pycolmap.essential_matrix_from_pose(j_from_i)
     l2d_j = to_homogeneous(p2d_i) @ j_E_i.T
     l2d_i = to_homogeneous(p2d_j) @ j_E_i
     dist = np.abs(np.sum(to_homogeneous(p2d_i) * l2d_i, axis=1))

hloc/utils/geometry.py

@@ -1,9 +1,10 @@
 from pathlib import Path
-from typing import Tuple
+from typing import Mapping, Tuple
 
 import cv2
 import h5py
 import numpy as np
+import pycolmap
 
 from .parsers import names_to_pair, names_to_pair_old
 
@@ -76,3 +77,16 @@ def get_matches(path: Path, name0: str, name1: str) -> Tuple[np.ndarray]:
         matches = np.flip(matches, -1)
     scores = scores[idx]
     return matches, scores
+
+
+def write_poses(
+    poses: Mapping[str, pycolmap.Rigid3d], path: str, prepend_camera_name: bool
+):
+    with open(path, "w") as f:
+        for query, t in poses.items():
+            qvec = " ".join(map(str, t.rotation.quat[[3, 0, 1, 2]]))
+            tvec = " ".join(map(str, t.translation))
+            name = query.split("/")[-1]
+            if prepend_camera_name:
+                name = query.split("/")[-2] + "/" + name
+            f.write(f"{name} {qvec} {tvec}\n")

hloc/utils/io.py

@@ -116,7 +116,7 @@ def plot_camera(
             legendgroup=legendgroup,
             name=name,
             showlegend=False,
-            hovertemplate=text.replace("\n", "<br>"),
+            hovertemplate=text.replace("\n", "<br>") if text else None,
         )
         fig.add_trace(pyramid)
 
@@ -134,25 +134,37 @@ def plot_camera(
         name=name,
         line=dict(color=color, width=1),
         showlegend=False,
-        hovertemplate=text.replace("\n", "<br>"),
+        hovertemplate=text.replace("\n", "<br>") if text else None,
     )
     fig.add_trace(pyramid)
 
 
 def plot_camera_colmap(
-    fig: go.Figure,
-    image: pycolmap.Image,
-    camera: pycolmap.Camera,
-    name: Optional[str] = None,
-    **kwargs
+    fig: go.Figure, cam_from_world: pycolmap.Rigid3d, camera: pycolmap.Camera, **kwargs
 ):
     """Plot a camera frustum from PyCOLMAP objects"""
-    world_t_camera = image.cam_from_world.inverse()
+    world_t_camera = cam_from_world.inverse()
     plot_camera(
         fig,
         world_t_camera.rotation.matrix(),
         world_t_camera.translation,
         camera.calibration_matrix(),
+        **kwargs
+    )
+
+
+def plot_image_colmap(
+    fig: go.Figure,
+    image: pycolmap.Image,
+    camera: pycolmap.Camera,
+    name: Optional[str] = None,
+    **kwargs
+):
+    """Plot a camera frustum from a PyCOLMAP image."""
+    plot_camera_colmap(
+        fig,
+        image.cam_from_world(),
+        camera,
         name=name or str(image.image_id),
         text=str(image),
         **kwargs
@@ -162,9 +174,7 @@ def plot_camera_colmap(
 def plot_cameras(fig: go.Figure, reconstruction: pycolmap.Reconstruction, **kwargs):
     """Plot a camera as a cone with camera frustum."""
     for image_id, image in reconstruction.images.items():
-        plot_camera_colmap(
-            fig, image, reconstruction.cameras[image.camera_id], **kwargs
-        )
+        plot_image_colmap(fig, image, reconstruction.cameras[image.camera_id], **kwargs)
 
 
 def plot_reconstruction(
@@ -186,8 +196,7 @@ def plot_reconstruction(
         p3D
         for _, p3D in rec.points3D.items()
         if (
-            (p3D.xyz >= bbs[0]).all()
-            and (p3D.xyz <= bbs[1]).all()
+            bbs.contains_point(p3D.xyz)
             and p3D.error <= max_reproj_error
             and p3D.track.length() >= min_track_length
         )