🎬 Discover the Power of DLC Model Zoo in the GUI ! 🚀

This spring the DLC team released the code and preprint on a new method, called, SuperAnimal, to build plug-and-play deep learning models that can be used immediately on common animals🐄🐅🦒 without additional model training🏋️‍♂️.

• To learn more about the tech behind the SuperAnimal Models check out our previous blogpost !
• You can also read the preprint HERE for more technical details!

Here we cover how to use the SuperAnimal models via the DLC GUI✨!

➡️ Let’s GO!

First make sure that your DLC version is up-to-date. Simply, activate your DLC conda environment and launch the GUI.

You should have the DeepLabCut Project Manager GUI 2.3.3 (or newer)!

💡Notice that in the bottom right corner there is a Model Zoo🐿🐘🐭 button now!

To update your DeepLabCut code:

1. In terminal/Anaconda prompt activate your DLC conda environment

conda activate DEEPLABCUT

2. Run the update command

pip install --upgrade deeplabcut

Excellent✨! Your DeepLabCut environment is now up-to-date👍.

Before starting, create a video folder and place the 🎥 video you wish to process inside. The 📁video folder can be located anywhere.

For this example, we’ll create a folder on the desktop called “mouse” and use a short top-view mouse video from Sturman et al.

The reference video featuring the mouse comes from (2)

➡️ Go ahead and click the Model Zoo button.

Behold the Model Zoo GUI🔥🔥!

The video selection area allows you to upload a video and specify its file format.

Model Zoo currently supports three file formats:

1️⃣ mp4

2️⃣ mov

3️⃣ avi

If you uploaded the wrong video, you can easily clear your selection and choose a new one.

Once done, proceed to the Supermodel settings.

You can choose from 2️⃣ available SuperAnimal models:

• Superanimal Quadruped 🦌: Perfect for analyzing quadrupeds from a side view!
• Superanimal topviewmouse 🐭: Specially crafted for lab mice in different lab settings, as seen from a bird’s-eye 🦅 view!

Select the appropriate model for your video from the dropdown menu and hit RUN!

DLC will work its magic now.

Sit back, relax, and enjoy the processing progress in the terminal window! 🍿

Once completed, head back to your video folder and examine 🔎 the results.

The content should include:
1️⃣ Your original video 📽️

The reference video featuring the mouse comes from (2)

2️⃣ A labeled version, enriched with model insights! ✨

3️⃣ H5 files containing pose estimation tracking key points, both before and after video adaptation, ready for in-depth analysis 🎯

4️⃣ A plot-poses folder filled with:

• Valuable tracking data 📈
• A visual representation of key points the model is tracking 🔍

If your tracking involve subjects not covered by the SuperAnimal models or you require a customized project, fear not 🙌 ! It’s easy to set up. Simply return to the main page, click on ‘create new project,’ and follow the step-by-step guidance tailored to your unique requirements.

For more information on custom projects and other features, our previous Medium blog post on GUI usage serves as an informative and helpful resource.

Don’t forget to take a look at our video 📽 demonstration on YouTube to see the GUI Model Zoo🦓🐅🐘 in action!

#DeepLabCut

Awesome work 🔥

References📚

1. Ye, S. et al. (2023), SuperAnimal models pretrained for plug-and-play analysis of animal behavior, arXiv.org. Available at: https://arxiv.org/abs/2203.07436
2. Sturman, O., von Ziegler, L., Schläppi, C. et al. Deep learning-based behavioral analysis reaches human accuracy and is capable of outperforming commercial solutions.Neuropsychopharmacol.45, 1942–1952 (2020). https://doi.org/10.1038/s41386-020-0776-y