166 Reply by joan

(2 replies, posted in General)

Do I understand correctly that you are *producing* new kva files out of excel, not the other way around?

The mapping from timestamps to frames changes with each video, it's decided by the encoder. But there is some logic in the kva importer to adapt files created from one video when loaded into another with different frame rate. Can you clarify your process a bit more, how is the excel sheet created, is it referencing times or frames, do you do any annotations in Kinovea or is it another program that creates this excel sheet that you want to import, etc.

Go to forum: General

167 Reply by joan

(3 replies, posted in Cameras and hardware)

Apart from here there are some occasional discussions on the Golf Simulator forum, especially the Computer Systems sub-forum.

Maybe of interest also the Aravis discussion board, Aravis is an open source library to work with GenICam machine vision cameras. The forum is of course focused on the use of the library but you might find some interesting info there. (note: this library requires changing the driver of the cameras to the generic libusb one which typically makes them not compatible with the manufacturers software).

For instance when just looking at the information I can find on for instance the IMX287 I can't see any indication that I can get more fps at lower resolution but things I have read indicates that it should be possible.

Yeah I think that's because it's not a feature of the sensor itself but of the firmware. Most likely all sensors are capable of this, even those in webcams, action cams or phones, but the on board chip and its firmware driving the sensor are exposing different functionalities.

Many vendors have an area on their website with "white papers" or a "knowledge base" or "application notes", with interesting info at the level you might be interested in, not necessarily specific to one particular camera.

Some links:

- https://www.baslerweb.com/en/downloads/documents/
- https://www.baslerweb.com/en/learning/
- https://en.ids-imaging.com/knowledge-base.html
- https://www.baumer.com/us/en/service-su … w-Overview
- https://www.teledynevisionsolutions.com … edge-base/

Go to forum: Cameras and hardware

168 Reply by joan

(3 replies, posted in Bug reports)

How are the files created? What video player can read them? Could you share a small sample file in a new issue in github (click Bugs at the bottom of this page) or by email joan at kinovea dot org. Thanks.

Go to forum: Bug reports

169 Reply by joan

(3 replies, posted in General)

The threshold setting is just above the replacement frame rate, in Preferences > Capture > Recording.
I don't think it's related to the problem, it only changes a metadata entry in the output file. The file will contain the same frames.

Retroactive is a recording mode, it's on the same preference page as replacement frame rate. It will record the video to the memory buffer, and only at the end of the recording will generate the output file on disk. The other methods write to the disk on the fly which can be slower (but can be used continuously for longer). See more details here: https://www.kinovea.org/help/en/userint … #recording

Exposure time is in the individual camera configuration dialog (from the wrench button in the lower left of the capture screen). Make sure it's lower than 4 milliseconds if you target 250 fps. https://www.kinovea.org/help/en/capture … ebcam.html

Go to forum: General

170 Reply by joan

(3 replies, posted in General)

> Why is replacement frame rate maxed at 100fps?

The point of this feature is to create a file with a traditional (low) framerate inside it, so it can still be played back normally (although at slower speed than real time). So starting with a high speed camera but saving files with normal speed frame rate and video in slow motion (typical operating mode of a high speed camera). To disable this feature you can set the "Threshold" field higher than your actual capture frame rate so it will never trigger. Maybe there should be an explicit check box to make things clearer.

The limited number of frames in the output video and the blur seems to be two other separate problems. Blur is caused by too high exposure time. What camera model are you using?
Try to use the recording mode "Retroactive" for best performance if you are capturing short clips that fit in the delay buffer.

Go to forum: General

171 Reply by joan

(1 replies, posted in Bug reports)

Thanks, very interesting. Unfortunately I'm afraid there is not much I can do about this right now but it confirms that the component used in the file explorer is too old for modern Windows and needs to be replaced.

Go to forum: Bug reports

172 Reply by joan

(14 replies, posted in General)

ahan wrote:

Can we add an operation method like keyframe and RV to click and drag in the center of the video to watch frame by frame. There is also the function of frame extraction playback. Thank you, it would be perfect if there were such an operational method.

Hold the ALT key and click and drag the video, it will move frame by frame (or mouse wheel back and forth).

I don't know what you mean by "frame extraction playback", please describe with more details. If you mean playback of image sequences, it is already supported (for the common image formats, not EXR).

By RV I expect you mean Autodesk RV or Open RV, I'm not familiar with this program.

Go to forum: General

173 Reply by joan

(4 replies, posted in Bug reports)

The behavior was changed to make it more robust and more standard conformant.

I don't have Excel at hand here but when I try in online Excel in OneDrive, it fails by default and puts everything in a single column, but there is a "Paste options" button with a "Split Text to Columns" option where I can select the correct variant. (edit: I can also do this after the paste by selecting the cells and doing Data > Split Text to Columns).

Are you saying that the version without the quotes is splitting the columns by default correctly on your machine and the one with the quotes is not?

The quotes should really not change anything. They are necessary if the text can contain the decimal separator which is the case here since the headers are user-entered text.

When I paste in LibreOffice, if I keep the headers, it automatically shows the text import dialog. Then the values are imported as text or as numbers based on the decimal separator.

If I paste in Google Sheets it's the same as OneDrive Excel, fails by default but there is a button "paste formatting" with a "Split text to columns" option with the options.

"Time (ms)";"Trajectory 1"
"133";"142.0046"
"167";"150.0579"

"Time (ms)";"Trajectory 1"
"133";"142,0046"
"167";"150,0579"

Time (ms);Trajectory 1
133;142.0046
167;150.0579

Time (ms);Trajectory 1
133;142,0046
167;150,0579

Go to forum: Bug reports

174 Reply by joan

(4 replies, posted in Bug reports)

The double quotes don't make the content string, it's a standard way of writing the data in case the cell separator is the same as the decimal separator.

What is the problem you are facing exactly? Is there a spreadsheet software that won't take this and create numbers? Does it not go through a text import dialog?

I'm surprised by the fact that the first value seemingly uses a coma and the others a point as decimal separator. You can force the separator in preferences under Drawings > Export.

Go to forum: Bug reports

175 Reply by joan

(4 replies, posted in General)

There are no PM here. And it would be great to have these sorts of discussions in the open so other people can read them and learn through them.

Go to forum: General

176 Reply by joan

(5 replies, posted in Bug reports)

Very interesting. So maybe it's the file explorer thing that's blocking everything for certain people. This part of the code is based on some truly ancient Windows APIs.

Do you have network drives bound to drive letters? Or some other slightly atypical things that might trip it up?

Go to forum: Bug reports

177 Reply by joan

(5 replies, posted in Bug reports)

There have been several similar reports over the years but unfortunately I've never been able to reproduce it, so I don't know what to investigate. I have tested on several Windows 11 machines including a fresh install and never got the problem.

Go to forum: Bug reports

178 Reply by joan

(1 replies, posted in General)

I'm not sure if this would work but assuming there are landmarks on the rail tracks or pylons that are equally spaced, one way might be to use the counter tool and mark one beat each time the pylon gets out of view.

Right now the counter tool data doesn't get exported in the spreadsheet export… that's an oversight. You could do the same adding a full key frame at each landmark crossing event. Then in the spreadsheet you would need to multiply the frame index by the distance between the pylons and that should give you a distance/time series that you could do further calculations on.

Go to forum: General

179 Reply by joan

(1 replies, posted in Bug reports)

Thanks, I reproduce the problem. I'm not sure what's going on.

Go to forum: Bug reports

180 Reply by joan

(5 replies, posted in General)

I think the most pressing issue is that when tracking drawings we don't see what's happening and there is no way to configure the tracking parameters. This has been a problem for too long and it's time to address it.

This is my current tentative roadmap/priorities regarding tracking improvements (these may or may not make sense, more personal notes than anything)

Roadmap
1. Move tracking configuration to a new tab in the side panel.
2. Have a visual feedback of the tracking always visible in this panel.
3. Have an option to keep the trajectory alive and recover when tracking fails for a few frames (occlusions).
4. Make it possible to configure and visualize the parameters of each tracked point of drawings in this same panel.
5. Make it possible to display the individual trajectories of the tracked points of drawings.
6. Implement new tracking algorithms for round markers and for "quadrants" markers (2x2 checker board pattern).
7. Implement new style option to see the trajectories as a series of points.
8. Implement "onion skinning" for segments of tracked drawings, to see multiple positions at once.

Now a system to attach trackable points of a drawing to existing trajectories would come after this, depending on feedback. At that point the purpose of this would be to support the workflow of tracking individual points one after the other instead of all at once. But I feel the quest for this workflow is mainly a consequence of the current system not being robust nor user friendly enough.

I think points 4 and 5 will alleviate most of the pain of tracking drawings.


I also want to distinguish (at least in my mind) between the two contexts in which we do tracking.

1. Tracking for visual feedback (qualitative)
- typically without any well defined plane of motion or impossible to calibrate.
- typically markerless.
- may or may not have camera motion.
- trajectories might be very complex in 3D space.
- lots of occlusions and unclear positions, motion blur.

Use cases: pretty much everything filmed in the field, tv footage, golf/disc golf shot tracing, track and field, gymnastics, dance, etc.
Tools: pattern matching, ball tracking, human pose estimation, spline tool.

2. Tracking for measurement
- typically with a very well defined plane of motion perpendicular to the camera axis.
- joints positions identified by markers.
- static camera.
- 2D motion only.
- limited occlusions.

Use cases: gait analysis, posture analysis, biomechanics studies, bike fit.
Tools: round marker tracking, quadrants marker tracking, 6DoF tracking.

Go to forum: General