Making sense of the log files and the console outputs

[nikunjsanghai]

Hi,
I have read the paper and was trying to recreate the results under Fig 2 and Fig 3 talking about 4 different motions. First random(Fig 2) then Fig(3) deals with circular, forward and lateral motions. Please correct me if I am wrong. The issue I am having is that when I run the main_expectation.py file, the objective scores for different cameras is different from what is mentioned in the paper. According to my understanding the current main file is giving me 2 to 6 k(camera positions) for it with objective scores, I think the median RMSE values I am getting on the console seem almost same, but the objective scores i got are as follows:
Best Score till now: 1.1915899979021663e-12
Next best Camera is:
R: [
6.12323e-17, 0, 1;
0, 1, 0;
-1, 0, 6.12323e-17
]
t: 0.15 0 0

---

Best Score till now: 2.061258456434215
Next best Camera is:
R: [
-0.5, 0, 0.866025;
0, 1, 0;
-0.866025, 0, -0.5
]
t: -0.15 0 -0.15

---

Best Score till now: 5.720280523043846
Next best Camera is:
R: [
-1.83697e-16, 0, -1;
0, 1, 0;
1, 0, -1.83697e-16
]
t: 0.15 0 -0.15

---

Best Score till now: 6.001388289750302
Next best Camera is:
R: [
-0.866025, 0, -0.5;
0, 1, 0;
0.5, 0, -0.866025
]
t: -0.15 0 0.15

---

Best Score till now: 6.037328144844131
Next best Camera is:
R: [
0.866025, 0, -0.5;
0, 1, 0;
0.5, 0, 0.866025
]
t: -0.15 0 0

---

Best Score till now: 6.049854050133662
Next best Camera is:
R: [
-1, 0, 1.22465e-16;
0, 1, 0;
-1.22465e-16, 0, -1
]
t: 0.15 0 0.15

---

Selected candidates are :
[[ 5]
[23]
[33]
[37]
[43]
[67]]
The score for traj greedy: 6.049854050

I am trying to recreate the simulation results first with an end goal of verifying the results with new data and getting similar results. Any help would be appreciated. I am not able to make sense of the objective scores here and why am I getting a run from k=2 to k=6 cameras here where as in the main file it states
if name == 'main':
''' construct the 3D world and the trajectory'''
''' Sample all the camera configurations. In sim I have ~300 configs '''
''' The goal is to pick the best N among these placeents.'''
''' Run greedy first, get a initial baseline.'''
''' Use greedy solution as initial value'''

parser = argparse.ArgumentParser(formatter_class=argparse.RawTextHelpFormatter,
                                 description='runs experiments for different benchmark \
                                             algorithms for optimal camera placement\n\n')

parser.add_argument('-n', '--num_runs', help='number of runs in the experiment', default=10)
parser.add_argument('-s', '--select_k', help='number of cameras to select', default=2)
parser.add_argument('-t', '--traj_type', help='Type of trajectory 1:circle, 2:side, 3:forward, 4:random', default=4)
parser.add_argument('-o', '--output_dir', help='Output dir for output bag file', default='.')
parser.add_argument('-c', '--config_file',
                    help='Yaml file which specifies the topic names, frequency of selection and time range',
                    default='config/config.yaml')

so shouldn't I get only the 2 camera placements since the default is 2 and I have not provided specific arguments. When I am looking at the log files, there seem to be 30, for 10 runs of each equal, standard, and random. My understanding is that these are the 3 benchmark algorithms the paper was talking about to be used for comparision to show how greedy and frank-wolfe differ from the 3.