Trajectory file format

[chraibi]

In Gitlab by @chraibi on Sep 25, 2017, 17:18 [origin]

Currently only trajectories files with the unit "cm" are supported.

[chraibi]

In Gitlab by @gjaeger on Jan 29, 2018, 21:54

for example in txt2txt.py

[chraibi]

In Gitlab by @chraibi on Jan 29, 2018, 21:59

assigned to @gjaeger

[chraibi]

In Gitlab by @chraibi on Jan 29, 2018, 22:01

changed time estimate to 2d

[chraibi]

In Gitlab by @gjaeger on Jan 30, 2018, 09:54

changed title from {-Support unit cm-} to {+scripts | txt2txt.py+}

[chraibi]

In Gitlab by @gjaeger on Jan 30, 2018, 09:54

changed the description

[chraibi]

In Gitlab by @gjaeger on Jan 30, 2018, 10:30

https://gitlab.version.fz-juelich.de/jupedsim/jpsreport/commit/cdff542fcfa088bd76ac95cf3d24d2476ef69cb3 : After a short discussion with Jette: It would be better to convert a particular file than all txt-files in a folder. Previously, there was a danger to edit already edited files.

[chraibi]

In Gitlab by @chraibi on Jan 30, 2018, 12:55

Yes I think so. Sometimes it can be annoying..

What about we have two options?

• -f filename: parse filename
• -p directory: parse all files in directory

[chraibi]

In Gitlab by @gjaeger on Jan 31, 2018, 11:38

PeTrack offers eleven export options. How do we deal with that?

Files can be very different. We could pass the number and names of columns through a list. Isn't it easier to limit the script to one option and parse one file?

[chraibi]

In Gitlab by @chraibi on Jan 31, 2018, 12:42

What are these 11 options?

[chraibi]

In Gitlab by @gjaeger on Jan 31, 2018, 12:52

from PeTrack 0.8

[chraibi]

In Gitlab by @chraibi on Jan 31, 2018, 12:56

Ok, how do these options influence the resulting trajectory files?
Is then the format different?

[chraibi]

In Gitlab by @gjaeger on Jan 31, 2018, 13:02

Yes.

For example: experiment Corridor, unidirectional flow

This is the first line (UNI_CORR_500_9.txt)

# <number> <frame> <x> [in cm] <y> [in cm] <z> [in cm] <rot> [in rad] <id> <flag>

We should find out which export options are written to the txt-file.

[chraibi]

In Gitlab by @boltes1 on Feb 1, 2018, 17:34

The additional numbers " [in rad] " are comming from an extraction programm from Wolfgang Mehner; especially the id is interesting while using the individual information of each person.

[chraibi]

In Gitlab by @gjaeger on Oct 9, 2018, 21:37

extract from export function in PeTrack:

void TrackerReal::exportTxt(QTextStream &out, bool alternateHeight, bool useTrackpoints, bool exportViewingDirection, bool exportAngleOfView, bool exportUseM, bool exportMarkerID)
{
   float scale;

   out << "# z: can be 3d position or height of person (alternating or not)" << endl;
   if (exportViewingDirection)
       out << "# viewDirX viewDirY: vector of direction of head of person" << endl;
   if (exportAngleOfView)
       out << "# viewAngle: angle of view of camera to person from perpendicular [0..Pi/2]" << endl;
   if (exportUseM)
       out << "# id frame x/m y/m z/m";
   else
       out << "# id frame x/cm y/cm z/cm";
   if (exportViewingDirection)
       out << " viewDirX viewDirY";
   if (exportAngleOfView)
       out << " viewAngle" << endl;
   if (exportMarkerID)
       out << " markerID" << endl;
   else
       out << endl;
... 
}

[chraibi]

In Gitlab by @gjaeger on Oct 9, 2018, 21:38

Currently we have the following column names:

JPScore (txt-file)

• ID (person/agent ID)
• FR (frame)
• X (x-position)
• Y (y-position)
• Z (z-position)
• rA (radius A of elipse)
• rB (radius A of elipse)
• eO (elipseOrientation)
• eC (elipseColor)
• e0 (desired direction)

PeTrack (txt-file)

• id (person ID)
• frame (frame)
• x/m (x-position - alternative: x/cm)
• y/m (y-position - alternative: y/cm)
• z/m (z-position - alternative: z/cm)
• viewDirX (x-vector of direction of head of person)
• viewDirY (y-vector of direction of head of person)
• viewAngle (angle of view of camera to person from perpendicular [0 .. Pi/2])
• markerID

@chraibi @boltes1:
My suggestion for a consolidated column name list for trajectory-files:

• PersID (person/agent ID)
• Frame (frame)
• x/m (x-position)
• y/m (y-position)
• z/m (z-position)
• V (speed of pedestrian in m/s)
• Vx (x component of the pedestrian's velocity)
• Vy (y component of the pedestrian's velocity)
• Vz (z component of the pedestrian's velocity)
• rA (radius A of elipse)
• rB (radius A of elipse)
• eO (elipseOrientation)
• eC (elipseColor)
• desired direction
  • Dx (x component of the pedestrian's desired direction)
  • Dy (y component of the pedestrian's desired direction)
• viewDirX (x-vector of direction of head of person)
• viewDirY (y-vector of direction of head of person)
• viewAngle (angle of view of camera to person from perpendicular [0 .. Pi/2])
• markerID

Edit: sorry, I forgot the column markerID in the first version.
Edit: Minor additions

[chraibi]

In Gitlab by @gjaeger on Oct 10, 2018, 07:55

We should add a version of the file specification to the header.

Where can we describe the file specification? On the website jupedsim.org or ped.fz-juelich.de/database?

[chraibi]

In Gitlab by @chraibi on Apr 6, 2019, 17:58

Sorry, I forgot about this.
We should talk about it next time with the PeTracks colleagues.

[chraibi]

In Gitlab by @gjaeger on Apr 6, 2019, 18:01

I'm prepared. 😄

[chraibi]

In Gitlab by @boltes1 on Apr 6, 2019, 19:00

eine Vereinheitlichung wird schwer, da je nach Experiment andere Dinge drinstehen; einzig "id, frame, x, y" ist immer gleich und drin;
fuer das Projekt SiME wurden u.a. Kommentare fuer individuelle Eigenschaften hinzugefuegt, die im Kopf der Datei aufgefuehrt sind;
was aber wenigstens fuer die neuen Dateien gewaehrleistet ist ist, dass im Haeder beschrieben steht, was die Datei beinhaltet;
in Zukunft kann ich mir beim Export des 3D-Skeletts noch ganz andere Dinge vorstellen;
auch in Jupedsim wird es doch nicht bei den Ellipsen bleiben;
gerne koennen wir aber noch einmal darueber reden

[chraibi]

In Gitlab by @gjaeger on Apr 8, 2019, 09:16

I can understand your objections.

I extend my suggestion: The development could be considered by versioning the list of column names. This ensures that no column name is used twice. In a documentation you could explain the column name in more detail. The version of the column name and the version of the program should be listed in the header.

Versioning would be a step for research data management. With the introduction of a versioning, we can continue to use old data in the future. No knowledge would be lost - hopefully. This allows us to document our data even better.

[chraibi]

In Gitlab by @gjaeger on May 12, 2019, 22:11

some hints for me:

jpscore - new trajectory format based on issue jpscore - issue JuPedSim/jpscore#310

[chraibi]

In Gitlab by @gjaeger on May 22, 2019, 10:07

from issue #107:

We need a column name for body height (head height (AH) and we should specify the description of the z-position (z/m).

[gjaeger]

I have extended my proposal. The following was added:

• V (speed of pedestrian in m/s)
• Vx (x component of the pedestrian's velocity)
• Vy (y component of the pedestrian's velocity)
• Vz (z component of the pedestrian's velocity)

[gjaeger]

I've extended my proposal. The following was added:

• desired direction
  • Dx (x component of the pedestrian's desired direction)
  • Dy (y component of the pedestrian's desired direction)