Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Task04 Игорь Логинов SPbU #26

Closed
wants to merge 3 commits into from
Closed

Task04 Игорь Логинов SPbU #26

wants to merge 3 commits into from

Conversation

I-7
Copy link

@I-7 I-7 commented Jun 13, 2024 •
edited
Loading

  1. test_ceres_solver/FitLine: почему найденная прямая и эталонная - не совпадают? Как это исправить пост-обработкой? Как это исправить формулировкой задачи?

Прямые не совпадают т.к. при умножении коэффициентов Ax+By+C=0 на константу прямая остаётся той же. Можно либо как-то нормализовать коэффициенты на этапе пост-обработки (например, сделать A=1, если A!=0; иначе сделать B=0). Либо эти же требования на коэффициенты вынести в формулировку задачи.

  1. BA: представьте что вы написали преобразование phg::Calibration -> блок параметров и обратное блок параметров -> phg::Calibration. Как проверить простым образом что эти преобразования сделаны корректно? Что должно быть в логе про процент inliers до/после BA если runBA() вызывать всегда два раза пордяд? Иначе говоря - что следует из того что в идеале runBA() должна быть (мне очень нравится это слово) - идемпотентна?

Повторный вызов runBA() должен в идеале ничего не менять, но надёжнее проверять, что изменения от второго вызова значительно меньше, чем от первого - какие-то коррекции всё-таки могут произойти (по аналогии с тем, что в конце градиентного спуска мы должны стоять на месте, но на самом деле будем с каждым шагом смещаться всё меньше и меньше).

  1. Какое максимальное число кадров у вас получилось хорошо выравнять для каждого из датасетов? (проверьте хотя бы saharov32 и herzjesu25) Не забудьте приложить скриншоты.

20 для saharov32 (остановился на этом, т.к. дальше камеры ещё дольше приклеиваются, а это заняло ~час на моей виртуалке)

image

аналогично 20 для herzjesu25

image

  1. Если бы вычисления в double были абсолютно точны - можно ли было бы назвать вычисления в Calibration::project/unproject строго зеркальными?

В project у нас "хорошее" r, а в unproject мы знаем только оценку на r (по x и y).

  1. Почему фокальная длина меняется от того что мы уменьшаем картинку? Почему именно f/downscale?

Мы сжали картинки = уменьшили мир = уменьшили фокальную длину в одинаковое число раз.

  1. Имеет ли право BA двигать точку отсчета системы координат (т.е. добавить константу ко всем координатам)? Как это повлияет на суммарную Loss?

Идейно мы можем это делать (сдвинуть все точки и камеры на одинаковый векторы) и вроде это не должно никак влиять на loss...

  1. Каким образом можно гарантировать чтобы при сравнении нескольких последовательно построенных облаков точек одного и того же датасета (созданных по мере добавления фотографии за фотографией) в MeshLab - облака не были хаотично смещены/отмасштабированы/повернуты друг от друга?

Кажется, что у нас первая камера всегда в (0, 0, 0) и учитывая специфику задачи последовательные облака точек будут почти совпадать.

Тут облако точек по 16 камерам одним цветом, по 20 - другим:

image

CI

Running main() from /home/runner/work/PhotogrammetryTasks2024/PhotogrammetryTasks2024/libs/3rdparty/libgtest/googletest/src/gtest_main.cc
[==========] Running 5 tests from 1 test suite.
[----------] Global test environment set-up.
[----------] 5 tests from CeresSolver
[ RUN      ] CeresSolver.HelloWorld1
iter      cost      cost_change  |gradient|   |step|    tr_ratio  tr_radius  ls_iter  iter_time  total_time
   0  1.250000e+01    0.00e+00    5.00e+00   0.00e+00   0.00e+00  1.00e+04        0    1.10e-05    [7](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:8).41e-05
   1  1.249750e-07    1.25e+01    5.00e-04   5.00e+00   1.00e+00  3.00e+04        1    3.10e-05    4.4[8](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:9)e-04
   2  1.388518e-16    1.25e-07    1.67e-08   5.00e-04   1.00e+00  [9](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:10).00e+04        1    5.96e-06    4.66e-04
Ceres Solver Report: Iterations: 3, Initial cost: 1.250000e+01, Final cost: 1.388518e-16, Termination: CONVERGENCE
x:     5 -> [10](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:11)
f(x):  5 -> 1.66644e-08
f'(x): -1 -> -1
[       OK ] CeresSolver.HelloWorld1 (1 ms)
[ RUN      ] CeresSolver.HelloWorld2
iter      cost      cost_change  |gradient|   |step|    tr_ratio  tr_radius  ls_iter  iter_time  total_time
   0  6.131250e+04    0.00e+00    1.40e+04   0.00e+00   0.00e+00  1.00e+04        0    3.81e-06    2.79e-05
   1  3.127313e+04    3.00e+04    7.82e+03   4.47e+01   4.90e-01  1.00e+04        1    9.06e-06    4.91e-05
   2  5.685772e+02    3.07e+04    7.75e+02   1.01e+02   9.82e-01  3.00e+04        1    6.91e-06    6.70e-05
   3  1.137952e+02    4.55e+02    2.90e+02   2.55e+01   8.00e-01  3.83e+04        1    5.01e-06    8.20e-05
   4  1.931689e+00    1.12e+02    3.82e+01   1.97e+01   9.83e-01  1.15e+05        1    5.96e-06    9.70e-05
   5  6.620880e-01    1.27e+00    2.23e+01   4.56e+00   6.57e-01  1.18e+05        1    5.01e-06    1.12e-04
   6  1.715778e-02    6.45e-01    2.45e+00   4.10e+00   9.74e-01  3.55e+05        1    5.01e-06    1.26e-04
   7  2.040825e-03    1.51e-02    8.69e-01   8.05e-01   8.81e-01  6.37e+05        1    5.01e-06    1.40e-04
   8  3.955432e-04    1.65e-03    1.23e-01   8.19e-01   8.06e-01  8.28e+05        1    5.01e-06    1.54e-04
   9  4.204592e-05    3.53e-04    3.09e-02   1.49e-01   8.94e-01  1.62e+06        1    5.96e-06    1.70e-04
  10  1.366023e-05    2.84e-05    5.85e-03   1.57e-01   6.75e-01  1.69e+06        1    2.10e-05    2.00e-04
  [11](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:12)  1.455687e-06    1.22e-05    7.13e-04   2.79e-02   8.93e-01  3.29e+06        1    5.01e-06    2.15e-04
  [12](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:13)  4.821191e-07    9.74e-07    7.85e-04   2.95e-02   6.69e-01  3.43e+06        1    5.96e-06    2.30e-04
  [13](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:14)  5.128650e-08    4.31e-07    2.64e-04   5.24e-03   8.94e-01  6.69e+06        1    5.01e-06    2.44e-04
  [14](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:15)  1.696961e-08    3.43e-08    1.71e-04   5.54e-03   6.69e-01  6.96e+06        1    5.01e-06    2.58e-04
  [15](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:16)  1.803689e-09    1.52e-08    5.60e-05   9.82e-04   8.94e-01  1.36e+07        1    5.01e-06    2.72e-04
  16  5.966373e-10    1.21e-09    3.29e-05   1.04e-03   6.69e-01  1.41e+07        1    5.01e-06    2.87e-04
  17  6.339313e-11    5.33e-10    1.07e-05   1.84e-04   8.94e-01  2.77e+07        1    5.96e-06    3.08e-04
  18  2.096924e-11    4.24e-11    6.19e-06   1.95e-04   6.69e-01  2.88e+07        1    5.01e-06    3.23e-04
  19  2.227619e-12    1.87e-11    2.02e-06   3.45e-05   8.94e-01  5.62e+07        1    4.05e-06    3.33e-04
  20  7.368654e-13    1.49e-12    1.[16](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:17)e-06   3.65e-05   6.69e-01  5.85e+07        1    5.01e-06    3.49e-04
  21  7.827280e-14    6.59e-13    3.79e-07   6.47e-06   8.94e-01  1.14e+08        1    5.01e-06    3.62e-04
  22  2.589[18](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:19)8e-14    5.24e-14    2.18e-07   6.84e-06   6.69e-01  1.[19](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:20)e+08        1    4.77e-06    3.78e-04
Ceres Solver Report: Iterations: 23, Initial cost: 6.131250e+04, Final cost: 2.589188e-14, Termination: CONVERGENCE
Found intersection point: (10, 5, [20](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:21)0)
[       OK ] CeresSolver.HelloWorld2 (0 ms)
[ RUN      ] CeresSolver.FitLineNoise
iter      cost      cost_change  |gradient|   |step|    tr_ratio  tr_radius  ls_iter  iter_time  total_time
   0  4.506931e+07    0.00e+00    1.61e+08   0.00e+00   0.00e+00  1.00e+04        0    1.08e-04    1.79e-04
   1  4.634902e+02    4.51e+07    1.33e+04   1.00e+02   1.00e+00  3.00e+04        1    1.98e-04    3.90e-04
   2  4.631061e+02    3.84e-01    1.61e-01   5.13e-03   1.00e+00  9.00e+04        1    1.72e-04    5.74e-04
Ceres Solver Report: Iterations: 3, Initial cost: 4.506931e+07, Final cost: 4.631061e+02, Termination: CONVERGENCE
Found line: (a=1, b=-1.99972, c=200.071)
[       OK ] CeresSolver.FitLineNoise (1 ms)
[ RUN      ] CeresSolver.FitLineNoiseAndOutliers
iter      cost      cost_change  |gradient|   |step|    tr_ratio  tr_radius  ls_iter  iter_time  total_time
   0  4.776667e+07    0.00e+00    1.62e+08   0.00e+00   0.00e+00  1.00e+04        0    1.03e-04    1.52e-04
   1  1.784980e+06    4.60e+07    1.44e+06   1.01e+02   1.01e+00  3.00e+04        1    1.64e-04    3.29e-04
   2  1.780933e+06    4.05e+03    8.75e+03   4.62e-01   1.01e+00  9.00e+04        1    1.68e-04    5.07e-04
Ceres Solver Report: Iterations: 3, Initial cost: 4.776667e+07, Final cost: 1.780933e+06, Termination: CONVERGENCE
Found line: (a=1, b=-1.97298, c=198.695)
[       OK ] CeresSolver.FitLineNoiseAndOutliers (1 ms)
[ RUN      ] CeresSolver.FitLineNoiseAndOutliersWithHuberLoss
iter      cost      cost_change  |gradient|   |step|    tr_ratio  tr_radius  ls_iter  iter_time  total_time
   0  7.706573e+05    0.00e+00    1.40e+06   0.00e+00   0.00e+00  1.00e+04        0    1.07e-04    1.55e-04
   1  1.829705e+04    7.52e+05    9.32e+05   9.96e+01   2.00e+00  3.00e+04        1    1.71e-04    3.38e-04
   2  1.614312e+04    2.15e+03    3.09e+04   6.08e-02   1.12e+00  9.00e+04        1    1.58e-04    5.07e-04
   3  1.6141[23](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:11:24)e+04    1.89e+00    1.07e+02   2.82e-03   1.00e+00  2.70e+05        1    1.63e-04    6.79e-04
Ceres Solver Report: Iterations: 4, Initial cost: 7.706573e+05, Final cost: 1.614123e+04, Termination: CONVERGENCE
Found line: (a=1, b=-1.99961, c=200.072)
[       OK ] CeresSolver.FitLineNoiseAndOutliersWithHuberLoss (1 ms)
[----------] 5 tests from CeresSolver (5 ms total)

[----------] Global test environment tear-down
[==========] 5 tests from 1 test suite ran. (5 ms total)
[  PASSED  ] 5 tests.

Running main() from /home/runner/work/PhotogrammetryTasks2024/PhotogrammetryTasks2024/libs/3rdparty/libgtest/googletest/src/gtest_main.cc
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from SFM
[ RUN      ] SFM.ReconstructNViews
32 images
detecting points...
matching points...
1% - Cameras 0-1 (IMG_3023.JPG-IMG_3024.JPG): 1159 matches
4% - Cameras 3-0 (IMG_3026.JPG-IMG_3023.JPG): 180 matches
6% - Cameras 0-2 (IMG_3023.JPG-IMG_3025.JPG): 442 matches
8% - Cameras 6-1 (IMG_3029.JPG-IMG_3024.JPG): 3[7](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:8) matches
9% - Cameras 3-1 (IMG_3026.JPG-IMG_3024.JPG): 40[8](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:9) matches
10% - Cameras 0-3 (IMG_3023.JPG-IMG_3026.JPG): 148 matches
12% - Cameras 6-2 (IMG_302[9](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:10).JPG-IMG_3025.JPG): 61 matches
13% - Cameras 3-2 (IMG_3026.JPG-IMG_3025.JPG): 1229 matches
14% - Cameras 0-4 (IMG_3023.JPG-IMG_3027.JPG): 53 matches
16% - Cameras 8-3 (IMG_3031.JPG-IMG_3026.JPG): 21 matches
17% - Cameras 6-3 (IMG_3029.JPG-IMG_3026.JPG): 289 matches
18% - Cameras 3-4 (IMG_3026.JPG-IMG_3027.JPG): 1526 matches
20% - Cameras 8-4 (IMG_3031.JPG-IMG_3027.JPG): 182 matches
21% - Cameras 6-4 (IMG_3029.JPG-IMG_3027.JPG): 896 matches
22% - Cameras 3-5 (IMG_3026.JPG-IMG_3028.JPG): 765 matches
24% - Cameras 8-5 (IMG_3031.JPG-IMG_3028.JPG): 361 matches
26% - Cameras 6-5 (IMG_3029.JPG-IMG_3028.JPG): 1668 matches
27% - Cameras 3-6 (IMG_3026.JPG-IMG_3029.JPG): 268 matches
29% - Cameras 8-6 (IMG_3031.JPG-IMG_3029.JPG): 876 matches
30% - Cameras 6-7 (IMG_3029.JPG-IMG_3030.JPG): 1549 matches
31% - Cameras 3-7 (IMG_3026.JPG-IMG_3030.JPG): 80 matches
33% - Cameras 8-7 (IMG_3031.JPG-IMG_3030.JPG): 1254 matches
34% - Cameras 6-8 (IMG_3029.JPG-IMG_3031.JPG): 805 matches
38% - Cameras 8-9 (IMG_3031.JPG-IMG_3032.JPG): 1556 matches
39% - Cameras 6-9 (IMG_3029.JPG-IMG_3032.JPG): 469 matches
40% - Cameras 1-0 (IMG_3024.JPG-IMG_3023.JPG): [10](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:11)21 matches
44% - Cameras 1-2 (IMG_3024.JPG-IMG_3025.JPG): [11](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:12)45 matches
46% - Cameras 4-0 (IMG_3027.JPG-IMG_3023.JPG): 78 matches
49% - Cameras 4-1 (IMG_3027.JPG-IMG_3024.JPG): 148 matches
50% - Cameras 1-3 (IMG_3024.JPG-IMG_3026.JPG): 324 matches
53% - Cameras 4-2 (IMG_3027.JPG-IMG_3025.JPG): 653 matches
54% - Cameras 1-4 (IMG_3024.JPG-IMG_3027.JPG): [12](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:13)8 matches
57% - Cameras 7-3 (IMG_3030.JPG-IMG_3026.JPG): 100 matches
59% - Cameras 4-3 (IMG_3027.JPG-IMG_3026.JPG): 1678 matches
60% - Cameras 9-4 (IMG_3032.JPG-IMG_3027.JPG): 77 matches
61% - Cameras 7-4 (IMG_3030.JPG-IMG_3027.JPG): 311 matches
62% - Cameras 4-5 (IMG_3027.JPG-IMG_3028.JPG): [13](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:14)80 matches
64% - Cameras 9-5 (IMG_3032.JPG-IMG_3028.JPG): 289 matches
66% - Cameras 7-5 (IMG_3030.JPG-IMG_3028.JPG): 999 matches
67% - Cameras 4-6 (IMG_3027.JPG-IMG_3029.JPG): 1052 matches
69% - Cameras 9-6 (IMG_3032.JPG-IMG_3029.JPG): 454 matches
70% - Cameras 7-6 (IMG_3030.JPG-IMG_3029.JPG): 1571 matches
71% - Cameras 4-7 (IMG_3027.JPG-IMG_3030.JPG): 263 matches
73% - Cameras 9-7 (IMG_3032.JPG-IMG_3030.JPG): 738 matches
74% - Cameras 7-8 (IMG_3030.JPG-IMG_3031.JPG): 1199 matches
76% - Cameras 4-8 (IMG_3027.JPG-IMG_3031.JPG): [14](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:15)5 matches
78% - Cameras 9-8 (IMG_3032.JPG-IMG_3031.JPG): 1587 matches
79% - Cameras 7-9 (IMG_3030.JPG-IMG_3032.JPG): 731 matches
80% - Cameras 4-9 (IMG_3027.JPG-IMG_3032.JPG): 129 matches
81% - Cameras 2-0 (IMG_3025.JPG-IMG_3023.JPG): 509 matches
83% - Cameras 2-1 (IMG_3025.JPG-IMG_3024.JPG): 1[15](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:16)7 matches
84% - Cameras 5-1 (IMG_3028.JPG-IMG_3024.JPG): 20 matches
86% - Cameras 2-3 (IMG_3025.JPG-IMG_3026.JPG): 1256 matches
87% - Cameras 5-2 (IMG_3028.JPG-IMG_3025.JPG): 149 matches
88% - Cameras 2-4 (IMG_3025.JPG-IMG_3027.JPG): 648 matches
89% - Cameras 5-3 (IMG_3028.JPG-IMG_3026.JPG): 850 matches
90% - Cameras 2-5 (IMG_3025.JPG-IMG_3028.JPG): 1[16](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:17) matches
91% - Cameras 5-4 (IMG_3028.JPG-IMG_3027.JPG): 1389 matches
92% - Cameras 2-6 (IMG_3025.JPG-IMG_3029.JPG): 56 matches
93% - Cameras 5-6 (IMG_3028.JPG-IMG_3029.JPG): [17](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:18)90 matches
94% - Cameras 2-7 (IMG_3025.JPG-IMG_3030.JPG): 38 matches
96% - Cameras 5-7 (IMG_3028.JPG-IMG_3030.JPG): 1025 matches
98% - Cameras 5-8 (IMG_3028.JPG-IMG_3031.JPG): 370 matches
100% - Cameras 5-9 (IMG_3028.JPG-IMG_3032.JPG): 336 matches
Initial alignment from cameras #0 and #1 (IMG_3023.JPG, IMG_3024.JPG)
Before BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Before BA projections: 89% inliers with MSE=1.71016
    Camera #0 projections: 89% inliers (1037/1159) with MSE=1.68797
    Camera #1 projections: 89% inliers (1035/1159) with MSE=1.73239
After BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
After BA tie poits: 0% old + 11% new = 11% total outliers
After BA projections: 89% inliers with MSE=0.36941
    Camera #0 projections: 89% inliers (1027/1159) with MSE=0.374369
    Camera #1 projections: 89% inliers (1028/1159) with MSE=0.364456
Append camera #2 (IMG_3025.JPG) to alignment via 792 common points
Before BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Before BA projections: 64% inliers with MSE=0.697122
    Camera #0 projections: 83% inliers (1095/1324) with MSE=0.392793
    Camera #1 projections: 77% inliers (1386/[18](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:19)10) with MSE=0.569545
    Camera #2 projections: 35% inliers (577/1639) with MSE=1.58111
After BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.[19](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:20)3329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.75432, 0.0435193, 0.691639] -> [-1.9332, 0.0213844, 0.574549]
After BA tie poits: 7% old + 7% new = 14% total outliers
After BA projections: 89% inliers with MSE=0.650715
    Camera #0 projections: 84% inliers (1109/1324) with MSE=0.98434
    Camera #1 projections: 89% inliers (1604/1810) with MSE=0.75[20](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:21)57
    Camera #2 projections: 93% inliers (1518/1639) with MSE=0.299898
Append camera #3 (IMG_3026.JPG) to alignment via 984 common points
Before BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Before BA projections: 77% inliers with MSE=0.90909
    Camera #0 projections: 84% inliers (1125/1346) with MSE=0.986453
    Camera #1 projections: 88% inliers (1638/1853) with MSE=0.755548
    Camera #2 projections: 80% inliers (1832/2293) with MSE=0.466373
    Camera #3 projections: 56% inliers (975/1738) with MSE=1.90963
After BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.9332, 0.0[21](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:22)3844, 0.574549] -> [-1.90091, 0.0101914, 0.499642]
Camera #3 center: [-2.70342, 0.0419379, 1.14095] -> [-2.73193, 0.0164815, 0.915656]
After BA tie poits: 10% old + 6% new = 16% total outliers
After BA projections: 86% inliers with MSE=0.407085
    Camera #0 projections: 81% inliers (1095/1346) with MSE=1.02692
    Camera #1 projections: 85% inliers (1570/1853) with MSE=0.532744
    Camera #2 projections: 88% inliers (2028/[22](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:23)93) with MSE=0.166646
    Camera #3 projections: 89% inliers (1551/1738) with MSE=0.156667
Append camera #4 (IMG_3027.JPG) to alignment via 1134 common points
Before BA camera: k1=0, k2=0, f=1585.5, cx=364, cy=546
Before BA projections: 87% inliers with MSE=0.512175
    Camera #0 projections: 81% inliers (1103/1354) with MSE=1.02263
    Camera #1 projections: 85% inliers (1581/1865) with MSE=0.530324
    Camera #2 projections: 88% inliers (2189/2475) with MSE=0.178846
    Camera #3 projections: 91% inliers ([23](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:24)83/2618) with MSE=0.164083
    Camera #4 projections: 85% inliers (2061/[24](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:25)17) with MSE=0.981577
After BA camera: k1=0, k2=0, f=1707.05, cx=319.351, cy=505.84
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.90091, 0.0101914, 0.499642] -> [-1.87127, 0.0185203, 0.528922]
Camera #3 center: [-2.73193, 0.0164815, 0.915656] -> [-2.65834, 0.0353923, 0.956179]
Camera #4 center: [-3.34626, 0.0488612, 1.5019] -> [-3.22971, 0.0757299, 1.54206]
After BA tie poits: 12% old + 5% new = 17% total outliers
After BA projections: 84% inliers with MSE=0.218122
    Camera #0 projections: 78% inliers (1053/1354) with MSE=0.396043
    Camera #1 projections: 82% inliers (1530/1865) with MSE=0.223[25](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:26)9
    Camera #2 projections: 85% inliers (2097/2475) with MSE=0.168933
    Camera #3 projections: 88% inliers (2315/[26](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:27)18) with MSE=0.195517
    Camera #4 projections: 83% inliers (2013/2417) with MSE=0.198384
Append camera #5 (IMG_3028.JPG) to alignment via 1132 common points
Before BA camera: k1=0, k2=0, f=1707.05, cx=319.351, cy=505.84
Before BA projections: 84% inliers with MSE=0.407035
    Camera #0 projections: 78% inliers (1053/1354) with MSE=0.396043
    Camera #1 projections: 82% inliers (1531/1867) with MSE=0.223282
    Camera #2 projections: 85% inliers (2108/2489) with MSE=0.171912
    Camera #3 projections: 88% inliers (2505/2834) with MSE=0.[27](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:28)1134
    Camera #4 projections: 85% inliers (2609/3055) with MSE=0.346992
    Camera #5 projections: 84% inliers (1733/2062) with MSE=1.14888
After BA camera: k1=0, k2=0, f=1649.2, cx=324.435, cy=515.111
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.87127, 0.0185203, 0.5[28](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:29)922] -> [-1.87569, 0.017992, 0.523627]
Camera #3 center: [-2.65834, 0.0353923, 0.956179] -> [-2.66811, 0.0345145, 0.943554]
Camera #4 center: [-3.2[29](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:30)71, 0.0757299, 1.54206] -> [-3.24627, 0.0752875, 1.52204]
Camera #5 center: [-3.75066, 0.113421, 2.25747] -> [-3.76232, 0.125518, 2.2418]
After BA tie poits: 14% old + 4% new = 17% total outliers
After BA projections: 83% inliers with MSE=0.210426
    Camera #0 projections: 77% inliers (1040/1354) with MSE=0.425028
    Camera #1 projections: 81% inliers (1511/1867) with MSE=0.226726
    Camera #2 projections: 83% inliers (2059/2489) with MSE=0.140356
    Camera #3 projections: 84% inliers (2373/2834) with MSE=0.152814
    Camera #4 projections: 84% inliers (2554/[30](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:31)55) with MSE=0.199767
    Camera #5 projections: 89% inliers (1826/2062) with MSE=0.243501
Append camera #6 (IMG_3029.JPG) to alignment via 1359 common points
Before BA camera: k1=0, k2=0, f=1649.2, cx=324.435, cy=515.111
Before BA projections: 84% inliers with MSE=0.490049
    Camera #0 projections: 77% inliers (1040/1354) with MSE=0.425028
    Camera #1 projections: 81% inliers (1518/1874) with MSE=0.22594
    Camera #2 projections: 83% inliers (2061/2491) with MSE=0.140267
    Camera #3 projections: 84% inliers (2388/2853) with MSE=0.152162
    Camera #4 projections: 84% inliers (2763/3299) with MSE=0.3[31](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:32)39
    Camera #5 projections: 89% inliers (2659/2975) with MSE=0.519156
    Camera #6 projections: 86% inliers (2279/2656) with MSE=1.5244
After BA camera: k1=-0.182379, k2=-0.0755031, f=1603.73, cx=319.013, cy=526.627
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193[32](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:33)9] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.87569, 0.017992, 0.523627] -> [-1.87679, 0.0224249, 0.516563]
Camera #3 center: [-2.66811, 0.0345145, 0.943554] -> [-2.68999, 0.0452095, 0.936282]
Camera #4 center: [-3.24627, 0.0752875, 1.52204] -> [-3.28844, 0.0948087, 1.51261]
Camera #5 center: [-3.76232, 0.125518, 2.2418] -> [-3.81854, 0.154179, 2.22006]
Camera #6 center: [-3.98059, 0.219396, 3.10378] -> [-4.05255, 0.247325, 3.06266]
After BA tie poits: 14% old + 4% new = 18% total outliers
After BA projections: 83% inliers with MSE=0.15851
    Camera #0 projections: 77% inliers (1036/1354) with MSE=0.[33](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:34)0605
    Camera #1 projections: 80% inliers (1503/1874) with MSE=0.196921
    Camera #2 projections: 82% inliers (20[34](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:35)/2491) with MSE=0.102412
    Camera #3 projections: 81% inliers (2321/2853) with MSE=0.103307
    Camera #4 projections: 81% inliers (2676/3299) with MSE=0.145901
    Camera #5 projections: 86% inliers (2562/2975) with MSE=0.153065
    Camera #6 projections: 87% inliers (2318/2656) with MSE=0.18176
Append camera #7 (IMG_3030.JPG) to alignment via 1634 common points
Before BA camera: k1=-0.182379, k2=-0.0755031, f=1603.73, cx=319.013, cy=526.627
Before BA projections: 84% inliers with MSE=0.[35](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:36)897
    Camera #0 projections: 77% inliers (1036/1354) with MSE=0.330605
    Camera #1 projections: 80% inliers (1503/1874) with MSE=0.196921
    Camera #2 projections: 82% inliers (2034/2491) with MSE=0.102412
    Camera #3 projections: 81% inliers (2322/2854) with MSE=0.103307
    Camera #4 projections: 81% inliers (2690/3324) with MSE=0.146576
    Camera #5 projections: 86% inliers (2762/3203) with MSE=0.163393
    Camera #6 projections: 88% inliers (2927/3314) with MSE=0.182586
    Camera #7 projections: 88% inliers (2318/2622) with MSE=1.66019
After BA camera: k1=-0.224441, k2=0.177686, f=1570.25, cx=320.195, cy=530.264
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.87679, 0.0224249, 0.516563] -> [-1.87885, 0.0228[36](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:37)9, 0.514165]
Camera #3 center: [-2.68999, 0.0452095, 0.936282] -> [-2.69545, 0.0466972, 0.931141]
Camera #4 center: [-3.28844, 0.0948087, 1.51261] -> [-3.29636, 0.0976875, 1.50419]
Camera #5 center: [-3.81854, 0.154179, 2.22006] -> [-3.8304, 0.158268, 2.20544]
Camera #6 center: [-4.05255, 0.247325, 3.06266] -> [-4.07084, 0.252108, 3.04009]
Camera #7 center: [-4.19901, 0.348391, 3.98343] -> [-4.23513, 0.35182, 3.95347]
After BA tie poits: 15% old + 3% new = 19% total outliers
After BA projections: 82% inliers with MSE=0.151222
    Camera #0 projections: 76% inliers (1032/1354) with MSE=0.344891
    Camera #1 projections: 80% inliers (1490/1874) with MSE=0.200273
    Camera #2 projections: 81% inliers (2013/2491) with MSE=0.0933167
    Camera #3 projections: 80% inliers (2272/2854) with MSE=0.0923299
    Camera #4 projections: 78% inliers (2599/3324) with MSE=0.125295
    Camera #5 projections: 83% inliers (2647/3203) with MSE=0.15287
    Camera #6 projections: 85% inliers (2818/3314) with MSE=0.152997
    Camera #7 projections: 89% inliers (2323/2622) with MSE=0.166476
Append camera #8 (IMG_3031.JPG) to alignment via 1472 common points
Before BA camera: k1=-0.224441, k2=0.177686, f=1570.25, cx=320.195, cy=530.264
Before BA projections: 81% inliers with MSE=0.352359
    Camera #0 projections: 76% inliers (1032/1354) with MSE=0.344891
    Camera #1 projections: 80% inliers (1490/1874) with MSE=0.200273
    Camera #2 projections: 81% inliers (2013/2491) with MSE=0.0933167
    Camera #3 projections: 80% inliers (2273/2855) with MSE=0.0924586
    Camera #4 projections: 78% inliers (2608/3336) with MSE=0.126107
    Camera #5 projections: 83% inliers (2671/3236) with MSE=0.154844
    Camera #6 projections: 82% inliers (2917/3541) with MSE=0.212289
    Camera #7 projections: 86% inliers (2701/3138) with MSE=0.239517
    Camera #8 projections: 79% inliers (1835/2319) with MSE=2.08398
After BA camera: k1=-0.125356, k2=-0.355677, f=1579.45, cx=328.982, cy=539.709
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.87885, 0.0228369, 0.514165] -> [-1.8777, 0.0191502, 0.509438]
Camera #3 center: [-2.69545, 0.0466972, 0.931141] -> [-2.69229, 0.0[37](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:38)9399, 0.922238]
Camera #4 center: [-3.29636, 0.0976875, 1.50419] -> [-3.29348, 0.0824011, 1.4922]
Camera #5 center: [-3.8304, 0.158268, 2.20544] -> [-3.829, 0.13541, 2.192[38](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:39)]
Camera #6 center: [-4.07084, 0.252108, 3.04009] -> [-4.07309, 0.221311, 3.02935]
Camera #7 center: [-4.23513, 0.35182, 3.95347] -> [-4.23581, 0.31222, 3.94883]
Camera #8 center: [-3.96423, 0.4308[39](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:40), 4.9913] -> [-3.95769, 0.403838, 4.99337]
After BA tie poits: 17% old + 3% new = 19% total outliers
After BA projections: 81% inliers with MSE=0.15961
    Camera #0 projections: 76% inliers (1028/1354) with MSE=0.397542
    Camera #1 projections: 79% inliers (1483/1874) with MSE=0.213727
    Camera #2 projections: 80% inliers (2002/2491) with MSE=0.0954307
    Camera #3 projections: 78% inliers (22[40](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:41)/2855) with MSE=0.0829349
    Camera #4 projections: 77% inliers (2553/3336) with MSE=0.127953
    Camera #5 projections: 80% inliers (2573/3236) with MSE=0.146713
    Camera #6 projections: 82% inliers (2914/35[41](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:42)) with MSE=0.149012
    Camera #7 projections: 86% inliers (2696/3138) with MSE=0.168496
    Camera #8 projections: 88% inliers (2040/2319) with MSE=0.206825
Append camera #9 (IMG_3032.JPG) to alignment via 1527 common points
Before BA camera: k1=-0.125356, k2=-0.355677, f=1579.45, cx=328.982, cy=539.709
Before BA projections: 80% inliers with MSE=0.332192
    Camera #0 projections: 76% inliers (1028/1354) with MSE=0.3975[42](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:43)
    Camera #1 projections: 79% inliers (1483/1874) with MSE=0.213727
    Camera #2 projections: 80% inliers (2002/2491) with MSE=0.095[43](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:44)07
    Camera #3 projections: 78% inliers (2240/2855) with MSE=0.0829349
    Camera #4 projections: 77% inliers (2554/3337) with MSE=0.127973
    Camera #5 projections: 80% inliers (2591/3258) with MSE=0.152518
    Camera #6 projections: 82% inliers (2935/3564) with MSE=0.153515
    Camera #7 projections: 86% inliers (2831/3293) with MSE=0.20322
    Camera #8 projections: 87% inliers (2773/3183) with MSE=0.353536
    Camera #9 projections: 72% inliers (1892/2642) with MSE=1.89577
After BA camera: k1=-0.154655, k2=-0.0963073, f=1561.2, cx=331.03, cy=5[44](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:45).917
Camera #0 center: [0, 0, 0] -> [0, 0, 0]
Camera #1 center: [-0.981112, 0.0066058, 0.193329] -> [-0.981112, 0.0066058, 0.193329]
Camera #2 center: [-1.8777, 0.0191502, 0.509438] -> [-1.87859, 0.0187318, 0.507978]
Camera #3 center: [-2.69229, 0.0379399, 0.922238] -> [-2.69511, 0.0372594, 0.920173]
Camera #4 center: [-3.29348, 0.0824011, 1.4922] -> [-3.2974, 0.0810715, 1.48958]
Camera #5 center: [-3.829, 0.13541, 2.19238] -> [-3.83358, 0.133243, 2.18814]
Camera #6 center: [-4.07309, 0.221311, 3.02935] -> [-4.07878, 0.217312, 3.02259]
Camera #7 center: [-4.23581, 0.31222, 3.94883] -> [-4.24423, 0.306547, 3.93805]
Camera #8 center: [-3.95769, 0.403838, 4.99337] -> [-3.97677, 0.39274, 4.96418]
Camera #9 center: [-3.58789, 0.475836, 5.87369] -> [-3.63062, 0.[45](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:46)468, 5.89855]
After BA tie poits: 17% old + 2% new = 18% total outliers
After BA projections: 81% inliers with MSE=0.152223
    Camera #0 projections: 76% inliers (1028/1354) with MSE=0.429194
    Camera #1 projections: 79% inliers (1480/1874) with MSE=0.225689
    Camera #2 projections: 80% inliers (1998/2491) with MSE=0.0963993
    Camera #3 projections: 78% inliers (2229/2855) with MSE=0.080657
    Camera #4 projections: 75% inliers (2515/3337) with MSE=0.115439
    Camera #5 projections: 78% inliers (2527/3258) with MSE=0.129568
    Camera #6 projections: 80% inliers (2851/3564) with MSE=0.152404
    Camera #7 projections: 84% inliers (2763/3293) with MSE=0.166[46](https://github.com/PhotogrammetryCourse/PhotogrammetryTasks2024/actions/runs/9510261012/job/26214443982#step:12:47)6
    Camera #8 projections: 88% inliers (2813/3183) with MSE=0.169964
    Camera #9 projections: 93% inliers (2459/2642) with MSE=0.126837
[       OK ] SFM.ReconstructNViews (23188 ms)
[----------] 1 test from SFM (23188 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (23188 ms total)
[  PASSED  ] 1 test.

@PolarNick239
Copy link
Member

Все хорошо, 9/10 баллов 👍 (т.к. после дедлайна)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@I-7 @PolarNick239