Comments (2)
pmoulon
commented on June 16, 2024
1
Hi. We don't have native support for descriptor of length 256. But you can match the point yourself and export the feature index as a ascii match file for each image pair and OpenMVG can digest that for ransac filtering.
from openmvg.
saatvikmah
commented on June 16, 2024
@pmoulon I'll explain what I'm trying to achieve. I have run Superpoint and Sphereglue (feature matching for spherical images) on my dataset and I have saved my matches file in the format of OpenMVG's matches.putative.txt file which is -
idx idx
num_matches
x y
x y
...
for the same dataset I've initialized the sfm_data.json file with the command -
openMVG_main_SfMInit_ImageListing -i <dataset_in> -o <matches_dir> -f 1 -c 7
After this when I try to run Geometric Filtering on the matches.putative.txt, following is my log -
INFO: [main_GeometricFilter.cpp:131] You called :
openMVG_main_GeometricFilter
--input_file: building/matches/sfm_data.json
--matches: building/matches/matches.putative.txt
--output_file: building/matches/matches.f.txt
Optional parameters:
--input_pairs
--output_pairs
--force false
--geometric_model a
--guided_matching 0
--cache_size unlimited
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 10%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 20%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 30%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 40%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 50%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 60%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 70%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 80%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 90%
INFO: [loggerprogress.hpp:79] [- Regions Loading -] 100%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 0%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 10%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 20%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 30%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 40%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 50%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 60%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 70%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 80%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 90%
INFO: [loggerprogress.hpp:79] [- Geometric filtering -] 100%
INFO: [graph_stats.hpp:53] Graph statistics:
#nodes: 385
#cc: 366
#singleton: 364
Node degree statistics: min: 1, max: 18, mean: 1, median: 1
INFO: [main_GeometricFilter.cpp:383] Task done in (s): 0
INFO: [main_GeometricFilter.cpp:386]
Export Adjacency Matrix of the pairwise's geometric matches
INFO: [main_SfM.cpp:157]
-----------------------------------------------------------
Structure from Motion:
-----------------------------------------------------------
INFO: [loggerprogress.hpp:79] [- Features Loading -] 10%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 20%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 30%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 40%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 50%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 60%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 70%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 80%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 90%
INFO: [loggerprogress.hpp:79] [- Features Loading -] 100%
INFO: [sequential_SfM.cpp:278] Track building
INFO: [sequential_SfM.cpp:281] Track filtering
INFO: [sequential_SfM.cpp:283] Track export to internal struct
INFO: [sequential_SfM.cpp:309]
------------------
-- Tracks Stats --
Tracks number: 95
Images Id:
37, 48, 62, 65, 66, 67, 68, 69, 106, 125, 165, 186, 190, 213, 218, 219, 221, 239, 272, 326, 345,
------------------
TrackLength, Occurrence
2 25
3 4
4 9
5 8
6 10
7 9
8 6
9 8
10 6
11 8
13 1
14 1
INFO: [loggerprogress.hpp:79] [Selection of an initial pair] 100%
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(65,67)
view basename:(frame-0066,frame-0068)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 0.703406
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 28
#residuals: 112
Initial RMSE: 32.8775
Final RMSE: 30.5582
Time (s): 0.0407901
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,68)
view basename:(frame-0068,frame-0069)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 0.628547
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 20
#residuals: 80
Initial RMSE: 21.1544
Final RMSE: 0.221932
Time (s): 0.0121849
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,219)
view basename:(frame-0068,frame-0220)
ERROR: [sequential_SfM.cpp:534] /!\ Robust estimation failed to compute E for this pair: {67,219}
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(37,67)
view basename:(frame-0038,frame-0068)
ERROR: [sequential_SfM.cpp:534] /!\ Robust estimation failed to compute E for this pair: {67,37}
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,69)
view basename:(frame-0068,frame-0070)
ERROR: [sequential_SfM.cpp:534] /!\ Robust estimation failed to compute E for this pair: {67,69}
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(66,67)
view basename:(frame-0067,frame-0068)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 0.357432
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0503 18:29:01.802279 1204707 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: Eigen failure. Unable to perform dense Cholesky factorization.
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 18
#residuals: 72
Initial RMSE: 1.55235
Final RMSE: 0.382708
Time (s): 0.0111651
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,272)
view basename:(frame-0068,frame-0273)
ERROR: [sequential_SfM.cpp:534] /!\ Robust estimation failed to compute E for this pair: {67,272}
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,218)
view basename:(frame-0068,frame-0219)
ERROR: [sequential_SfM.cpp:534] /!\ Robust estimation failed to compute E for this pair: {67,218}
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(165,186)
view basename:(frame-0166,frame-0187)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 1.98633
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 16
#residuals: 64
Initial RMSE: 89.2958
Final RMSE: 45.4725
Time (s): 0.009269
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,221)
view basename:(frame-0068,frame-0222)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 1.81115
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 20
#residuals: 80
Initial RMSE: 96.1378
Final RMSE: 75.8202
Time (s): 0.0112369
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,326)
view basename:(frame-0068,frame-0327)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 2.01414
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 6
#residuals: 24
Initial RMSE: 90.5349
Final RMSE: 2.74927
Time (s): 0.00549197
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:733]
SequentialSfMReconstructionEngine::ComputeResidualsMSE.
-- #Tracks: 1
-- Residual min: 0.0129674
-- Residual median: 0.457047
-- Residual max: 0.640964
-- Residual mean: 0.28509
INFO: [sequential_SfM.cpp:495] Putative starting pair info:
index:(67,326)
view basename:(frame-0068,frame-0327)
INFO: [sequential_SfM.cpp:538] Relative pose a-contrario upper_bound residual is: 2.01414
INFO: [sfm_data_BA_ceres.cpp:513]
Bundle Adjustment statistics (approximated RMSE):
#views: 2
#poses: 2
#intrinsics: 1
#tracks: 6
#residuals: 24
Initial RMSE: 90.5349
Final RMSE: 2.73924
Time (s): 0.00519514
--
Used motion prior: 0
INFO: [sequential_SfM.cpp:648]
=========================
MSE Residual InitialPair Inlier:
INFO: [sequential_SfM.cpp:733]
SequentialSfMReconstructionEngine::ComputeResidualsMSE.
-- #Tracks: 1
-- Residual min: 0.0312753
-- Residual median: 0.463654
-- Residual max: 0.589377
-- Residual mean: 0.285574
INFO: [sequential_SfM.cpp:932] -- Trying robust Resection of view: 218
INFO: [SfM_Localizer.cpp:333]
-------------------------------
-- Robust Resection statistics:
-- Resection status: 0
-- #Points used for Resection: 1
-- #Points validated by robust Resection: 0
-- Threshold: 0
-------------------------------
INFO: [sequential_SfM.cpp:733]
SequentialSfMReconstructionEngine::ComputeResidualsMSE.
-- #Tracks: 1
-- Residual min: 0.0312753
-- Residual median: 0.463654
-- Residual max: 0.589377
-- Residual mean: 0.285574
INFO: [sequential_SfM.cpp:235]
-------------------------------
-- Structure from Motion (statistics):
-- #Camera calibrated: 8 from 385 input images.
-- #Tracks, #3D points: 1
-------------------------------
Histogram of residuals:
0 | 2
0.0558 | 0
0.112 | 0
0.167 | 0
0.223 | 0
0.279 | 0
0.335 | 0
0.391 | 1
0.446 | 0
0.502 | 1
0.589
INFO: [main_SfM.cpp:618] Total Sfm took (s): 0
INFO: [main_SfM.cpp:620] ...Generating SfM_Report.html
INFO: [main_SfM.cpp:625] ...Export SfM_Data to disk.
INFO: [loggerprogress.hpp:79] [- Compute scene structure color -] 100%
I can't seem to get a point cloud generated after SfM has run for some reason. The matches file is the same as the one used by in-built feature matching.
My current hunch is that either the 256 vector length descriptor when used by SfM in the following code block is creating issues
switch(scene_initializer_enum)
{
case ESfMSceneInitializer::INITIALIZE_AUTO_PAIR:
OPENMVG_LOG_ERROR << "Not yet implemented.";
return EXIT_FAILURE;
break;
case ESfMSceneInitializer::INITIALIZE_MAX_PAIR:
scene_initializer.reset(new SfMSceneInitializerMaxPair(sfm_data,
feats_provider.get(),
matches_provider.get()));
break;
case ESfMSceneInitializer::INITIALIZE_EXISTING_POSES:
scene_initializer.reset(new SfMSceneInitializer(sfm_data,
feats_provider.get(),
matches_provider.get()));
break;
case ESfMSceneInitializer::INITIALIZE_STELLAR:
scene_initializer.reset(new SfMSceneInitializerStellar(sfm_data,
feats_provider.get(),
matches_provider.get()));
break;
default:
OPENMVG_LOG_ERROR << "Unknown SFM Scene initializer method";
return EXIT_FAILURE;
}
Also I noticed that there's no image_describer.json isn't being created when running custom methods, but its present when running SIFT. Is this file necessary for SfM to run successfully?
Another thing I noticed was that geometric filtering reduces the matched points from 9k-ish down to 500 which is a massive reduction. There is a high threshold set for super point already so its highly unlikely that these points are noise. Can we bypass the geometric filtering?
from openmvg.
Related Issues (20)
- the bug in main_ComputeFeatures HOT 1
- about Union-Find feature track conflicts HOT 1
- How to create pipeline in visualstudio. HOT 4
- SIFT_describer.hpp could't find sift.h HOT 2
- Incosistent result HOT 3
- ERROR: flag 'flagfile' was defined more than once HOT 6
- The main_sfm_location of camera orientation doubt. HOT 3
- How to test own dataset
- Integrating openMVG and openMVS HOT 1
- Some questions about positioning accuracy
- Error in executing SFM HOT 2
- error in 59% about openMVG_sample_describe_and_match_GUI HOT 5
- A question about loading image HOT 2
- cmake-gui select Visual Studio IDE and click configure. Getting stuck. HOT 1
- Stellar vs Global mode of construction HOT 1
- with CERES: 2.2.0, “openMVG_main_SfM” terminated abruptly HOT 1
- about reprojrction error and Residuals HOT 1
- What does “AUTO” mean in matching algorithm? HOT 3
- Problem in Gcp registration HOT 5
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from openmvg.