FAQs#

Before you start#

Can I use idtracker.ai on my videos?#

See our Example videos for the kind of videos idtracker.ai tracks well, and follow our Guidelines for good videos to get the best results.

Does idtracker.ai work on Windows?#

Yes. The Installation page includes instructions for Windows. We have tested idtracker.ai on Windows 10 and Windows 11.

Can I run idtracker.ai on a laptop?#

Yes. We run idtracker.ai, with all its features, on gaming laptops. Check the Requirements.

Can I use idtracker.ai if my computer does not have a good GPU?#

Yes, with some limitations. See Install without a graphics device.

Does idtracker.ai track single animals?#

Yes. Although idtracker.ai is designed to keep the identities of many animals, it also tracks videos with a single animal: set the number of animals to 1 in the Segmentation app. idtracker.ai then skips the identification steps, which are the ones that need a GPU. Tracking is faster, because it only needs to segment the video, and it runs well on computers without a GPU.

Can I track humans with idtracker.ai?#

We have not tried it. We expect idtracker.ai to track people well in videos recorded under laboratory conditions. Tracking people in natural environments (streets, parks…) is much harder, and idtracker.ai was not designed for it. Since idtracker.ai is free and open source, parts of its algorithm could be reused for that purpose.

Tracking#

Can I track many videos in batch, without the graphical interface?#

Yes. Save your parameters to a TOML file with Save parameters in the Segmentation app, then run:

idtrackerai --load parameters.toml --track

Call this command in a loop or a shell script to process many videos. You can override parameters for each video from the command line, for example --video_paths or --name. See Terminal usage.

How do I track a video that is split into several files?#

Select all the files in the Segmentation app, or list them in the video_paths parameter of the TOML file. idtracker.ai tracks them as one continuous video.

Can I track videos with several separate arenas?#

Yes. Draw one region of interest per arena in the Segmentation app. If the animals cannot move between arenas, also enable Exclusive Regions of Interest, so that each identity stays in its own arena.

Can idtracker.ai handle very long videos?#

Yes. Video length is only limited by your RAM and disk space. To track only part of a long video, use the tracking_intervals parameter.

How can I reduce memory usage during tracking?#

Lower the number_of_parallel_workers parameter (see Parallel processing). During segmentation, each worker can use up to 4 GB of RAM. To save disk space, set a stricter data_policy so that fewer intermediate files are kept (see Output).

After tracking#

How do I know if the tracking was successful?#

At the end of the Tracking log, idtracker.ai reports the estimated accuracy of the identification. The trajectory files are saved in session_[SESSION_NAME]/trajectories/ (see Output structure). To check the results visually, open the session in the Validator, which lists the frames with possible errors.

Does idtracker.ai give orientation and posture information?#

Not directly, but you can compute them:

  • The preprocessing/list_of_blobs.pickle file holds the contour and orientation of every blob (see Data analysis). Our Fish Midline script uses it to compute the nose, tail and midline of fish.

  • The Video Generator creates one cropped video per animal. Use these videos as input for pose estimation tools such as DeepLabCut or SLEAP.

Updates and problems#

How do I update idtracker.ai?#

Activate your Conda environment and run:

python -m pip install --upgrade idtrackerai

To update from version 4 or older, see Update idtracker.ai.

I found an error. What should I do?#

  1. Update idtracker.ai: the error may already be fixed in the latest version.

  2. For errors during installation, check the Installation Troubleshooting page.

  3. If the error persists, open an issue in our GitLab repository or email info@idtracker.ai. Attach the idtrackerai.log file (see Tracking log).