0) Setting up before class

RunwayML - make a user with your Umea email (important!) and make sure you can login to the software via the website. Alternatively you download the software in a desktop version here: https://runwayml.com/download/

Plan for today:

• [ ] Understanding the RunwayML interface
• [ ] Exploring different types of models:
  • [ ] Object Detection with Yolo
  • [ ] Text generation with GPT-2
  • [ ] Image generation with StyleGAN
• [ ] Training models

Expectations + debugging

<aside> 👉 You can follow along and run the same models as we do, or just observe and take notes.

• If you experience errors running a model try to restart it and/or refresh the browser tab.
• Have a bit of patience when you start a new model, it can take up to a few minutes before it has loaded and is actually running.
• Sometimes your model might not load while we demonstrate it, or it might load a lot faster than mine.
• Let us know verbally or in the chat if we are going to fast or too slow and we will try to adjust the pace. </aside>

1) Object detection with Yolo

YOLOv4: https://app.runwayml.com/models/runway/YOLOv4

Input type: Image 📷

Output Type: Array 📋 (Name of detected object, bounding box (x, y, w & h), confidence score)

<aside> 👉 Exercise: Run this model and show your web camera some different object you have on your desk. Notice how this is similar to the pre-trained models in ml5 we explored yesterday.

</aside>

2) Text generation models