AIBluefisher
rows = db.execute("SELECT * FROM images") Can you provide pre-processed dataset? please！
It's a wonderful and amazing job! Thank you for sharing the code. However, when pre-processing the data, I find that in DOGS\scripts\preprocess\meganerf_to_colmap.py line 126 "database_path = os.path.join(scene_dir, 'database.db')" , the 'database.db' is created. And in the following lines, 'database.db' is not populated with data. So I have the following problem: File "F:\DOGS\DOGS\scripts\preprocess\database.py", line 120, in fetch_images_from_database rows = db.execute("SELECT * FROM images") sqlite3.OperationalError: no such table: images I'm wondering if this is appropriate to ask for you pre-processed dataset or more details to pre-process dataset. Please!
Hi, I think you did not run COLMAP to create that database. Please follow here to compute camera poses and point clouds with COLMAP for your dataset.
@AIBluefisher Thanks a lot for your nice work.
I encountered the same issue when pre-processed for mill19.
Just to confirm, do we need to run COLMAP on mill19/{building, rubble} before executing python -m scripts.preprocess.meganerf_to_colmap?
I found this topic was already discussed in https://github.com/AIBluefisher/DOGS/issues/11.
@AIBluefisher Thanks a lot for your nice work.
I encountered the same issue when pre-processed for mill19.
Just to confirm, do we need to run COLMAP on mill19/{building, rubble} before executing python -m scripts.preprocess.meganerf_to_colmap?
Yes. You have to extract the feature points and correspondences before running this script. Since we need correspondences for triangulation with ground-truth camera poses provided by MegaNeRF.
Thanks a lot for the clarification!
I understand the following colmap operations are required:
- feature_extractor
- exhaustive_matcher
- point_triangulator