Resources & Downloads
Download all the files you need for this unit. Click the buttons below to save each file to your computer.
Downloads
How to Set Up Your Files
Step 1: Create Your Project Folder
- On your desktop, right-click and choose New > Folder.
- Name the folder
TeachableMachine_Unit. - Download all the files above and save them into this folder.
Step 2: Unzip the Dataset
- Find the
TeachableMachine_Demo_Datasets.zipfile in your folder. - Right-click on it and choose "Extract All..."
- Click Extract. This will create a new folder with all the image files inside.
- Inside you should see folders organised by lesson and shape type.
Step 3: Understand the Dataset Structure
The demo dataset is organised like this:
TeachableMachine_Demo_Datasets/
├── Balanced_Dataset/
│ ├── TRAIN/
│ │ ├── Circles/ (10+ images)
│ │ ├── Triangles/ (10+ images)
│ │ └── Squares/ (10+ images)
│ └── TEST/
│ ├── Circles/ (5 images)
│ ├── Triangles/ (5 images)
│ └── Squares/ (5 images)
│
└── Biased_Dataset/
├── Circles_YellowBG/ (circles on yellow)
└── Triangles_BlueBG/ (triangles on blue)
About the Datasets
Balanced Dataset vs Biased Dataset
Balanced dataset: Both classes have the same number of images, and all images use the same white background. The model has to learn the shape itself.
Biased dataset: Circles are on a yellow background and triangles are on a blue background. The model might learn the background colour instead of the shape — this is the "background trick" used in Lesson 2.
TRAIN vs TEST — Why It Matters
TRAIN folder: These images are used to teach the model. You upload them into Teachable Machine's classes.
TEST folder: These images are used to check the model. They must be images the model has never seen before. This tells you how the model will perform on new, real-world data.
If you test with training images, you might get 100% accuracy — but that doesn't mean the model actually learned. It might have just memorised. Always test with fresh images!
Teacher Notes: Managing Resources
- The demo dataset ZIP should be pre-downloaded onto student machines or available on a shared drive.
- If students struggle with unzipping, demonstrate it on the board first.
- For Lesson 2, students can use the Biased_Dataset folder directly, or draw their own.
- The balanced dataset is useful for Lessons 1, 3, and 4 as a fallback for students who haven't drawn enough images.