alexemg / dlc-cajal-course Goto Github PK

Thanks to all the feedback!

Already beforehand share:

• What minimal specs do you need for a computer if you want to use DLC (efficiently)?
• What limitations does COLAB have?
• Do you want to use the GUI / COLAB?
• already install DLC, read rel. papers
• share what you're interested in multi-animal / DLC live / ... / clustering / action segmentation / kinematics / neural data ..

Day 2 feedback:

• How to pick the most representative frames?
• What parameters are relevant for what?
• Can we train models for students over night (for students without GPU access)? --> mostly COLAB should be ideal?

Day 3 feedback:

• how to measure speed / benchmark runtime ?
• include further reading recommendations and refs in lecture page

Specific formatting feedback:

• #23
• one can link notebooks in a Jupyter book: https://jupyterbook.org/en/stable/content/executable/index.html --> (oner relevant notebook is: https://github.com/DeepLabCut/DeepLabCut/blob/master/examples/COLAB/COLAB_YOURDATA_TrainNetwork_VideoAnalysis.ipynb)
• also link to relevant time intervals of video tutorials for GUI

General problem solving:

• What have people struggled with? --> with solutions! (also link to the [DeepLabCut Forum(https://forum.image.sc/t/behavior-and-deeplabcut/23710) and Forum solutions). --> @KonradDanielewski
• Start FAQ section --> @KonradDanielewski
• Glossary of terms --> make section in resources @KonradDanielewski
• Link for prerequisites --> make section in resources @KonradDanielewski

Please add more aspects that I forgot?

We are looking forward to work with you!

Please check out the materials provided here. Ultimately, this will be a public Jupyter Book, just like the one of DeepLabCut!

For now, if you want to see it, you have to compile it locally with:

(first install): pip install -U jupyter-book
From the main folder, run the following command: jupyter-book build .
Then just open the index.html file in _build/html/index.html

To dos -- please help :)

There are quite a few TODO's in the document. They come in many flavors:

• TODO_TA are todos for all teaching assistants!
• TODO_AM are todos for me. You can help ;)
• more generally these personalized ones also exist for other people like Danbee... TODO_DK.

You can look for them by searching in GitHub, e.g. like this:

or of course in the cloned repository with any code editor.

Feedback?

Any other feedback is also welcome, please just open issues/PRs or discuss on discord or by email!

• link BUCTD video and mention DLC3
• update Modelzoo part -- GUI updates: DeepLabCut/DeepLabCut#2431
• update Cebra, #92
• final check

-> https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2760923/
-> Krakauer et al. 2017