fix merge conflict

.github/workflows/benchmark-self-hosted.yml

@@ -31,7 +31,7 @@ jobs:
     steps:
       - uses: actions/[email protected]
       - name: Cache frontend
-        uses: actions/cache@v3
+        uses: actions/cache@v4
         env:
           # Increase this value to reset cache
           CACHE_NUMBER_FRONTEND: 0

.github/workflows/benchmark.yml

@@ -38,7 +38,7 @@ jobs:
     steps:
       - uses: actions/[email protected]
       - name: Cache frontend
-        uses: actions/cache@v3
+        uses: actions/cache@v4
         env:
           # Increase this value to reset cache
           CACHE_NUMBER_FRONTEND: 0

.gitignore

@@ -188,6 +188,7 @@ output/
 
 # Cache dir
 cache/
+*.mat
 
 # Dev notebooks
 notebooks/

environment_linux.yml

@@ -50,7 +50,7 @@ dependencies:
       - opencv-python>=4.5.4.58
       - pydegensac
       - colour
-      - pycolmap>=0.1.0
+      - pycolmap==0.4.0
       - trimesh[easy]
       - gtsam==4.2
       - pydot

environment_linux_cpuonly.yml

@@ -50,7 +50,7 @@ dependencies:
       - opencv-python>=4.5.4.58
       - pydegensac
       - colour
-      - pycolmap>=0.1.0
+      - pycolmap==0.4.0
       - trimesh[easy]
       - gtsam==4.2
       - pydot

gtsfm/averaging/rotation/rotation_averaging_base.py

@@ -2,6 +2,7 @@
 
 Authors: Jing Wu, Ayush Baid
 """
+
 import abc
 import time
 from typing import Dict, List, Optional, Tuple
@@ -11,7 +12,7 @@
 from dask.delayed import Delayed
 from gtsam import Pose3, Rot3
 
-import gtsfm.utils.geometry_comparisons as comp_utils
+import gtsfm.utils.alignment as alignment_utils
 import gtsfm.utils.metrics as metric_utils
 from gtsfm.common.pose_prior import PosePrior
 from gtsfm.evaluation.metrics import GtsfmMetric, GtsfmMetricsGroup
@@ -106,7 +107,7 @@ def evaluate(self, wRi_computed: List[Optional[Rot3]], wTi_gt: List[Optional[Pos
         if len(wRi_computed) != len(wRi_gt):
             raise ValueError("Lengths of wRi_list and gt_wRi_list should be the same.")
 
-        wRi_aligned = comp_utils.align_rotations(wRi_gt, wRi_computed)
+        wRi_aligned = alignment_utils.align_rotations(wRi_gt, wRi_computed)
 
         metrics = []
         metrics.append(GtsfmMetric(name="num_rotations_computed", data=len([x for x in wRi_computed if x is not None])))

gtsfm/averaging/translation/averaging_1dsfm.py

@@ -10,6 +10,7 @@
 
 Authors: Jing Wu, Ayush Baid, Akshay Krishnan
 """
+
 import time
 import timeit
 from collections import defaultdict
@@ -35,7 +36,7 @@
 )
 
 import gtsfm.common.types as gtsfm_types
-import gtsfm.utils.geometry_comparisons as comp_utils
+import gtsfm.utils.alignment as alignment_utils
 import gtsfm.utils.logger as logger_utils
 import gtsfm.utils.metrics as metrics_utils
 import gtsfm.utils.sampling as sampling_utils
@@ -626,7 +627,7 @@ def compute_metrics(
             wTi_list.append(None)
         else:
             wTi_list.append(Pose3(wRi, wti))
-    wTi_aligned_list, _ = comp_utils.align_poses_sim3_ignore_missing(gt_wTi_list, wTi_list)
+    wTi_aligned_list, _ = alignment_utils.align_poses_sim3_ignore_missing(gt_wTi_list, wTi_list)
     wti_aligned_list = [wTi.translation() if wTi is not None else None for wTi in wTi_aligned_list]
     gt_wti_list = [gt_wTi.translation() if gt_wTi is not None else None for gt_wTi in gt_wTi_list]
 

gtsfm/bundle/bundle_adjustment.py

@@ -2,6 +2,7 @@
 
 Authors: Xiaolong Wu, John Lambert, Ayush Baid
 """
+
 import logging
 import time
 from collections import Counter
@@ -28,6 +29,7 @@
 )
 
 import gtsfm.common.types as gtsfm_types
+import gtsfm.utils.alignment as alignment_utils
 import gtsfm.utils.metrics as metrics_utils
 import gtsfm.utils.tracks as track_utils
 from gtsfm.common.gtsfm_data import GtsfmData
@@ -511,9 +513,9 @@ def evaluate(
             return ba_metrics
 
         # Align the sparse multi-view estimate after BA to the ground truth pose graph.
-        aligned_filtered_data = filtered_data.align_via_Sim3_to_poses(wTi_list_ref=poses_gt)
+        aligned_filtered_data = alignment_utils.align_gtsfm_data_via_Sim3_to_poses(filtered_data, wTi_list_ref=poses_gt)
         ba_pose_error_metrics = metrics_utils.compute_ba_pose_metrics(
-            gt_wTi_list=poses_gt, ba_output=aligned_filtered_data, save_dir=save_dir
+            gt_wTi_list=poses_gt, computed_wTi_list=aligned_filtered_data.get_camera_poses(), save_dir=save_dir
         )
         ba_metrics.extend(metrics_group=ba_pose_error_metrics)
 

gtsfm/common/gtsfm_data.py

@@ -3,6 +3,7 @@
 
 Authors: Ayush Baid, John Lambert, Xiaolong Wu
 """
+
 import itertools
 import logging
 from typing import Any, Dict, List, Optional, Tuple
@@ -12,7 +13,6 @@
 from gtsam import Pose3, SfmTrack, Similarity3
 
 import gtsfm.common.types as gtsfm_types
-import gtsfm.utils.geometry_comparisons as geometry_comparisons
 import gtsfm.utils.graph as graph_utils
 import gtsfm.utils.reprojection as reproj_utils
 
@@ -426,23 +426,6 @@ def filter_landmarks(self, reproj_err_thresh: float = 5) -> Tuple["GtsfmData", L
 
         return filtered_data, valid_mask
 
-    def align_via_Sim3_to_poses(self, wTi_list_ref: List[Optional[Pose3]]) -> "GtsfmData":
-        """Align estimated, sparse multiview result (self) to a set of reference poses.
-
-        Args:
-            wTi_list_ref: list of reference/target camera poses, ordered by camera index.
-
-        Returns:
-            aligned_data: sparse multiview result that is aligned to the poses above.
-        """
-        # these are the estimated poses (source, to be aligned)
-        wTi_list = self.get_camera_poses()
-        # align the poses which are valid (i.e. are not None)
-        # some camera indices may have been lost after pruning to largest connected component, leading to None values
-        # rSe aligns the estimate `e` frame to the reference `r` frame
-        _, rSe = geometry_comparisons.align_poses_sim3_ignore_missing(wTi_list_ref, wTi_list)
-        return self.apply_Sim3(aSb=rSe)
-
     def apply_Sim3(self, aSb: Similarity3) -> "GtsfmData":
         """Assume current tracks and cameras are in frame "b", then transport them to frame "a".
 

gtsfm/evaluation/compare_colmap_outputs.py

@@ -0,0 +1,124 @@
+"""Script to compare two reconstructions in Colmap's output format.
+
+Authors: Ayush Baid
+"""
+
+import argparse
+import os
+from typing import Dict, List
+
+import numpy as np
+import pycolmap
+from gtsam import Pose3, Similarity3
+from scipy.spatial.transform import Rotation
+
+import gtsfm.runner.gtsfm_runner_base as runner_base
+import gtsfm.utils.alignment as alignment_utils
+import gtsfm.utils.io as io_utils
+import gtsfm.utils.logger as logger_utils
+import gtsfm.utils.metrics as metric_utils
+from gtsfm.evaluation.metrics import GtsfmMetricsGroup
+
+logger = logger_utils.get_logger()
+
+
+def load_poses(colmap_dirpath: str) -> Dict[str, Pose3]:
+    wTi_list, img_fnames, _, _, _, _ = io_utils.read_scene_data_from_colmap_format(colmap_dirpath)
+
+    return dict(zip(img_fnames, wTi_list))
+
+
+def compare_poses(baseline_dirpath: str, eval_dirpath: str, output_dirpath: str) -> None:
+    """Compare the pose metrics between two reconstructions (Colmap format).
+
+    Args:
+        baseline_dirpath: Directory with baseline reconstruction.
+        current_dirpath: Directory with reconstruction which needs evaluation.
+        output_dirpath: Directory to save the metrics.
+    """
+    baseline_wTi_dict = load_poses(baseline_dirpath)
+    current_wTi_dict = load_poses(eval_dirpath)
+
+    common_fnames = baseline_wTi_dict.keys() & current_wTi_dict.keys()
+
+    if args.use_pycolmap_alignment:
+        current_reconstruction = pycolmap.Reconstruction(eval_dirpath)
+        baselineScurrent = Similarity3(
+            current_reconstruction.align_robust(
+                list(baseline_wTi_dict.keys()),
+                [wTi.translation() for wTi in baseline_wTi_dict.values()],
+                min_common_images=3,
+            ).matrix
+        )
+        current_wTi_dict = {fname: baselineScurrent.transformFrom(wTi) for fname, wTi in current_wTi_dict.items()}
+
+        aRb = baselineScurrent.rotation().matrix()
+        atb = baselineScurrent.translation()
+        rz, ry, rx = Rotation.from_matrix(aRb).as_euler("zyx", degrees=True)
+        logger.info("Sim(3) Rotation `aRb`: rz=%.2f deg., ry=%.2f deg., rx=%.2f deg.", rz, ry, rx)
+        logger.info(f"Sim(3) Translation `atb`: [tx,ty,tz]={str(np.round(atb,2))}")
+        logger.info("Sim(3) Scale `asb`: %.2f", float(baselineScurrent.scale()))
+
+    logger.info(
+        "Baseline: %d, current: %d , common: %d poses",
+        len(baseline_wTi_dict),
+        len(current_wTi_dict),
+        len(common_fnames),
+    )
+
+    baseline_wTi_list: List[Pose3] = []
+    current_wTi_list: List[Pose3] = []
+    for fname, wTi in baseline_wTi_dict.items():
+        baseline_wTi_list.append(wTi)
+        current_wTi_list.append(current_wTi_dict.get(fname))
+
+    if not args.use_pycolmap_alignment:
+        current_wTi_list, _ = alignment_utils.align_poses_sim3_ignore_missing(baseline_wTi_list, current_wTi_list)
+
+    i2Ui1_dict_gt = metric_utils.get_twoview_translation_directions(baseline_wTi_list)
+
+    wRi_aligned_list, wti_aligned_list = metric_utils.get_rotations_translations_from_poses(current_wTi_list)
+    baseline_wRi_list, baseline_wti_list = metric_utils.get_rotations_translations_from_poses(baseline_wTi_list)
+
+    metrics = []
+    metrics.append(metric_utils.compute_rotation_angle_metric(wRi_aligned_list, baseline_wRi_list))
+    metrics.append(metric_utils.compute_translation_distance_metric(wti_aligned_list, baseline_wti_list))
+    metrics.append(metric_utils.compute_relative_translation_angle_metric(i2Ui1_dict_gt, current_wTi_list))
+    metrics.append(metric_utils.compute_translation_angle_metric(baseline_wTi_list, current_wTi_list))
+
+    rotation_angular_errors = metrics[0]._data
+    translation_angular_errors = metrics[3]._data
+    metrics.extend(
+        metric_utils.compute_pose_auc_metric(
+            rotation_angular_errors, translation_angular_errors, save_dir=output_dirpath
+        )
+    )
+
+    ba_pose_metrics = GtsfmMetricsGroup(name="ba_pose_error_metrics", metrics=metrics)
+
+    runner_base.save_metrics_reports([ba_pose_metrics], metrics_path=output_dirpath)
+
+
+if __name__ == "__main__":
+    # Compare two reconstructions (in Colmap's output format). Right now, we just compare the poses.
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--baseline",
+        required=True,
+        help="Path to directory containing benchmark artifacts for the baseline",
+    )
+    parser.add_argument(
+        "--current",
+        required=True,
+        help="Path to directory containing benchmark artifacts for the current",
+    )
+    parser.add_argument("--output", required=True, help="Output for the json file for pose metrics")
+    parser.add_argument(
+        "--use_pycolmap_alignment", action="store_true", help="Use Pycolmap to align cameras between two reconstruction"
+    )
+    args = parser.parse_args()
+
+    os.makedirs(args.output, exist_ok=True)
+
+    compare_poses(args.baseline, args.current, args.output)

gtsfm/multi_view_optimizer.py

@@ -2,6 +2,7 @@
 
 Authors: Ayush Baid, John Lambert
 """
+
 from typing import Dict, List, Optional, Tuple
 
 import dask
@@ -13,6 +14,7 @@
 
 import gtsfm.common.types as gtsfm_types
 import gtsfm.utils.graph as graph_utils
+import gtsfm.utils.alignment as alignment_utils
 from gtsfm.averaging.rotation.rotation_averaging_base import RotationAveragingBase
 from gtsfm.averaging.translation.translation_averaging_base import TranslationAveragingBase
 from gtsfm.bundle.global_ba import GlobalBundleAdjustment
@@ -165,7 +167,7 @@ def create_computation_graph(
         ]
 
         # Align the sparse multi-view estimate before BA to the ground truth pose graph.
-        ba_input_graph = dask.delayed(ba_input_graph.align_via_Sim3_to_poses)(gt_wTi_list)
+        ba_input_graph = dask.delayed(alignment_utils.align_gtsfm_data_via_Sim3_to_poses)(ba_input_graph, gt_wTi_list)
 
         return ba_input_graph, ba_result_graph, viewgraph_two_view_reports_graph, multiview_optimizer_metrics_graph
 

gtsfm/runner/gtsfm_runner_base.py

@@ -288,7 +288,7 @@ def run(self) -> GtsfmData:
             local_cluster_kwargs = {
                 "n_workers": self.parsed_args.num_workers,
                 "threads_per_worker": self.parsed_args.threads_per_worker,
-                "dashboard_address": self.parsed_args.dashboard_port
+                "dashboard_address": self.parsed_args.dashboard_port,
             }
             if self.parsed_args.worker_memory_limit is not None:
                 local_cluster_kwargs["memory_limit"] = self.parsed_args.worker_memory_limit
@@ -414,9 +414,7 @@ def run(self) -> GtsfmData:
         return sfm_result
 
 
-def unzip_two_view_results(
-    two_view_results: Dict[Tuple[int, int], TWO_VIEW_OUTPUT]
-) -> Tuple[
+def unzip_two_view_results(two_view_results: Dict[Tuple[int, int], TWO_VIEW_OUTPUT]) -> Tuple[
     Dict[Tuple[int, int], Rot3],
     Dict[Tuple[int, int], Unit3],
     Dict[Tuple[int, int], np.ndarray],
@@ -436,7 +434,7 @@ def unzip_two_view_results(
         i2Ri1 = two_view_output[0]
         i2Ui1 = two_view_output[1]
         if i2Ri1 is None or i2Ui1 is None:
-            print(f"Skip {i1},{i2} since None")
+            logger.debug("Skip %d, %d since None", i1, i2)
             continue
 
         i2Ri1_dict[(i1, i2)] = i2Ri1

gtsfm/scene_optimizer.py

@@ -2,6 +2,7 @@
 
 Authors: Ayush Baid, John Lambert
 """
+
 import logging
 import os
 import shutil
@@ -14,23 +15,24 @@
 import matplotlib.pyplot as plt
 import numpy as np
 from dask.delayed import Delayed
-from gtsam import Pose3, Similarity3, Rot3, Unit3
+from gtsam import Pose3, Rot3, Similarity3, Unit3
 from trimesh import Trimesh
 
 import gtsfm.common.types as gtsfm_types
 import gtsfm.two_view_estimator as two_view_estimator
+import gtsfm.utils.alignment as alignment_utils
 import gtsfm.utils.ellipsoid as ellipsoid_utils
 import gtsfm.utils.io as io_utils
 import gtsfm.utils.logger as logger_utils
 import gtsfm.utils.viz as viz_utils
 from gtsfm.common.gtsfm_data import GtsfmData
 from gtsfm.common.image import Image
-from gtsfm.retriever.image_pairs_generator import ImagePairsGenerator
 from gtsfm.common.keypoints import Keypoints
 from gtsfm.common.pose_prior import PosePrior
 from gtsfm.densify.mvs_base import MVSBase
 from gtsfm.frontend.correspondence_generator.correspondence_generator_base import CorrespondenceGeneratorBase
 from gtsfm.multi_view_optimizer import MultiViewOptimizer
+from gtsfm.retriever.image_pairs_generator import ImagePairsGenerator
 from gtsfm.retriever.retriever_base import ImageMatchingRegime
 from gtsfm.two_view_estimator import (
     POST_ISP_REPORT_TAG,
@@ -297,8 +299,8 @@ def align_estimated_gtsfm_data(
         Updated ba_output GtsfmData object aligned to axes.
         Updated gt_pose_graph with GT poses aligned to axes.
     """
-    ba_input = ba_input.align_via_Sim3_to_poses(gt_wTi_list)
-    ba_output = ba_output.align_via_Sim3_to_poses(gt_wTi_list)
+    ba_input = alignment_utils.align_gtsfm_data_via_Sim3_to_poses(ba_input, gt_wTi_list)
+    ba_output = alignment_utils.align_gtsfm_data_via_Sim3_to_poses(ba_output, gt_wTi_list)
 
     walignedTw = ellipsoid_utils.get_ortho_axis_alignment_transform(ba_output)
     walignedSw = Similarity3(R=walignedTw.rotation(), t=walignedTw.translation(), s=1.0)

gtsfm/two_view_estimator_cacher.py

@@ -50,6 +50,8 @@ def __generate_cache_key(
         self, keypoints_i1: Keypoints, keypoints_i2: Keypoints, putative_corr_idxs: np.ndarray
     ) -> str:
         """Generates a cache key according to keypoint coordinates and putative correspondence indices."""
+        if putative_corr_idxs.size == 0:  # catch no correspondences
+            return cache_utils.generate_hash_for_numpy_array(np.array([]))
 
         # Subsample correspondence indices.
         sampled_idxs = putative_corr_idxs[:NUM_CORRESPONDENCES_TO_SAMPLE_FOR_HASH]