Request: Clarification on creating separate Conda environments for Grounded-SAM2, Hierarchical-Localization, and ROMP (like Metric3Dv2)

[vanamboy]

I want to create separate Conda environments for Grounded-SAM2, Hierarchical-Localization (hloc), and ROMP, similar to how Metric3Dv2 uses its own environment, to avoid package conflicts.

I noticed in the docs: SAM2_PYTHON_PATH = "/home/lixin/miniconda3/envs/sam21/bin/python"

Does this mean sam21 is a separate conda env for Grounded-SAM2?

Should I also create separate envs for hloc and ROMP?

Please confirm the recommended Python versions and if there are any naming conventions you'd suggest for these envs.

Thank you!

[lxxue]

Sorry for the confusion.

sam21 is the same environment as Grounded-SAM2, which should probably be renamed as hsr-data. I think you can install Gourned-SAM2 / hloc / ROMP in the same python environment.

Feel free to post more questions regarding environment setup and I will update the readme accordingly.

[vanamboy]

HSR-data-preprocessing: HLOC fails with AssertionError: matches.h5 — no pairs found for matching

I am running process_data.py with hloc, superpoint_aachen, and superglue using the sequential matching method. The process extracts features but reports Found 0 pairs and skips matching. This causes an AssertionError because matches.h5 is missing.

Here’s part of the log:
[hloc INFO] Found 0 pairs.
[hloc INFO] Matching local features with configuration:
{'model': {'name': 'superglue',
'sinkhorn_iterations': 50,
'weights': 'outdoor'},
'output': 'matches-superglue'}
[hloc INFO] Skipping the matching.
AssertionError: /home/vanamboy/Desktop/test/camera_poses/hloc-superpoint_aachen-sequential-superglue/matches.h5

I also see a warning about Grounded-SAM2's _C.so missing a symbol, but it says it's okay to ignore.

Kindly help me understand why no pairs were generated for matching and how I can fix this issue so that matches.h5 is created and HLOC works properly.

[vanamboy]

(/media/vanamboy/NewVolume/conda_envs/hsr-data) vanamboy@hp:/media/vanamboy/NewVolume/Test/HSR-data-preprocessing$ python process_data.py --data_dir /home/vanamboy/Desktop/test --input_path /media/vanamboy/NewVolume/Footage/shot_1 --gender male --sfm_tool hloc --feature_type superpoint_aachen --matcher_type superglue --matching_method sequential --camera_model PINHOLE --frame_start 1 --frame_end 9 --steps 0 1 2 3
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 9/9 [00:00<00:00, 46.06it/s]
Extracted 1 frames from 8 frames
FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
Generating human masks
frame loading (JPEG): 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 32.00it/s]
UserWarning: /media/vanamboy/NewVolume/Test/HSR-data-preprocessing/third_party/Grounded-SAM-2/sam2/_C.so: undefined symbol: _ZN3c106detail23torchInternalAssertFailEPKcS2_jS2_RKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

Skipping the post-processing step due to the error above. You can still use SAM 2 and it's OK to ignore the error above, although some post-processing functionality may be limited (which doesn't affect the results in most cases; see https://github.com/facebookresearch/sam2/blob/main/INSTALL.md).
propagate in video: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 503.34it/s]
Creating masks for colmap
[2025/06/19 13:53:04 hloc INFO] Extracting local features with configuration:
{'model': {'max_keypoints': 4096, 'name': 'superpoint', 'nms_radius': 3},
'output': 'feats-superpoint-n4096-r1024',
'preprocessing': {'grayscale': True, 'resize_max': 1024}}
Loaded SuperPoint model
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1.81it/s]
[2025/06/19 13:53:04 hloc INFO] Finished exporting features.
[2025/06/19 13:53:04 hloc INFO] Found 0 pairs.
[2025/06/19 13:53:04 hloc INFO] Matching local features with configuration:
{'model': {'name': 'superglue',
'sinkhorn_iterations': 50,
'weights': 'outdoor'},
'output': 'matches-superglue'}
[2025/06/19 13:53:04 hloc INFO] Skipping the matching.
Traceback (most recent call last):
File "/media/vanamboy/NewVolume/Test/HSR-data-preprocessing/estimate_camera_poses.py", line 242, in
run_hloc(
File "/media/vanamboy/NewVolume/Test/HSR-data-preprocessing/estimate_camera_poses.py", line 192, in run_hloc
reconstruction.main(
File "/media/vanamboy/NewVolume/Test/HSR-data-preprocessing/third_party/Hierarchical-Localization/hloc/reconstruction.py", line 122, in main
assert matches.exists(), matches
AssertionError: /home/vanamboy/Desktop/test/camera_poses/hloc-superpoint_aachen-sequential-superglue/matches.h5
Using cache found in /home/vanamboy/.cache/torch/hub/yvanyin_metric3d_main
/media/vanamboy/NewVolume/conda_envs/metric3d/lib/python3.10/site-packages/timm/models/layers/init.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)
/media/vanamboy/NewVolume/conda_envs/metric3d/lib/python3.10/site-packages/timm/models/registry.py:4: FutureWarning: Importing from timm.models.registry is deprecated, please import via timm.models
warnings.warn(f"Importing from {name} is deprecated, please import via timm.models", FutureWarning)
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:03<00:00, 3.13s/it]
(/media/vanamboy/NewVolume/conda_envs/hsr-data) vanamboy@hp:/media/vanamboy/NewVolume/Test/HSR-data-preprocessing$

[lxxue]

Because you only have 8 frames (idx 1 to 8), you should set --window_size 1 so that all the frames are used. The current run contains only one image, so there will be no matches found