Artificial intelligence → Module 6 → Final project

Final project: make your own AI project

Through the course you have learned what artificial intelligence is, how a machine learns, how AI is programmed and how it works on small devices. Now it's time to make something of your own. The project has three options – choose the one that matches what you know. All three are worth the same: what is graded is understanding, not how difficult the tools are.

Three options

The project is done alone or in pairs. In a pair, each person must be able to explain the whole project, not just their part.

🟢 Option 1: “AI around us” and Teachable Machine

Task: make a poster or a short presentation (6–8 slides) about artificial intelligence in your everyday life, together with your own small model in Teachable Machine.

1. AI around me: find 5 examples of AI that you or your family use (phone, recommendations, translator, navigation…). For each, write: what it does, what kind of data it probably learned from and what mistakes it can make.
2. My model: in Teachable Machine, make a model with 2–3 classes (for example objects from your desk, sounds, hand movements). Record at least 30 examples per class – objects, not people's faces.
3. Testing: try the model on 10 new examples and write down how accurate it was. Find one example where it makes a mistake and explain why.
4. Improvement: add examples or change the background and test again. Is it better?
5. Ethics: one slide (or part of the poster) about how such a model could be misused or be unfair, and how to prevent that.

Example topics: sorting waste (paper, plastic, metal); recognizing school supplies; applause, whistling and silence; “rock, paper, scissors” with hand movements.

🟡 Option 2: a classifier with your own data

Task: collect your own table of data, train a model in Python and check how good it is.

1. Question: choose what the model should predict and from which features. Examples: the type of leaf from its length and width; whether a day is “good for a walk” from the temperature, wind and cloudiness; the genre of a song from its tempo and length.
2. Data: measure and write down at least 40 examples in a .csv file (like fruit.csv from the lesson Getting ready). Write how you collected them.
3. Chart: draw the data (matplotlib) and describe whether the groups can be seen with the naked eye.
4. Model: split the data into training and test data (80% / 20%) and try at least two models – for example k-NN and a decision tree or a neural network. Compare the accuracy.
5. Error analysis: find the examples on which the model makes mistakes. Is data missing? Are some features unimportant?
6. Use: write a small program that asks the user for values and prints the prediction.

🔵 Option 3: a device that “learns”

Task: make a device that collects data with a sensor and makes decisions with a model trained on that data.

1. Idea: what does the device recognize? Examples: the state of a classroom or room (DHT11, LDR), whether a plant needs water (soil moisture sensor), hand movements (micro:bit), whether a door is open (ultrasonic sensor or accelerometer).
2. Collecting: the device sends measurements to the computer, as in the lesson ESP32 collects data. Collect at least 100 measurements and label them.
3. Learning: train a small decision tree in Python and move it to the board (A model from Python on the ESP32), or make gesture recognition on the micro:bit (micro:bit recognizes gestures).
4. Display: the device shows the decision – with an LED, a light ring, a screen or a web page.
5. Testing in reality: let the device run for at least 15 minutes and write down how many times its decision matched your own judgement.
6. What would you do differently: more data, a different sensor, different limits?

An idea for the whole class: the smart classroom from the Arduino and ESP32 course, where we no longer write the rules ourselves – a model learns them.

Work plan

The project log is the most important part: after each lesson, write down what you did, what didn't work and what you changed. The log shows that the project is yours.

Rules for using AI tools

• You may use AI tools as a helper – to explain an error or suggest an idea – but write in the log what you asked and what you used from it.
• Code you don't understand doesn't go into the project. At the presentation you must explain every part.
• Don't collect other people's personal data without their permission (and their parents' permission); don't record faces for the model.
• Mark images, text or sound made by AI as such.

More about this: How to use AI tools and Ethics and safety of AI.

Grading (100 points)

The scoring and the scale are a suggestion; the teacher can adjust them.

Checklist before handing in

• ☐ I can explain what my model predicts and from which data.
• ☐ I tested the model on examples it didn't see during training.
• ☐ I found at least one mistake of the model and explained it.
• ☐ I wrote how the project could be unfair or misused.
• ☐ The log says what I did and where I used AI tools.
• ☐ The project doesn't contain personal data or faces of other people without permission.

Common misconceptions

• “The more complex the model, the better the project” – a good project has a clear question, good data and a fair test – and a simple model can do that.
• “100% accuracy means the model is perfect” – probably the test was done on the same examples it learned from, or there are too few examples.
• “The model's mistakes should be hidden” – explained mistakes earn points – they show that you understand how the model works.

Practice tasks

Try it yourself first, and only then open the solution.

1. A student made a model that recognizes an apple and a pear from an image. In the test it had 100% accuracy, but at home the model keeps making mistakes. Give two possible reasons.
2. For option 🟡, suggest one topic: what the model predicts, which 2–3 features it uses and how you will collect 40 examples.
3. The project “Recognizing classmates by their faces”. What is the problem with this topic, and how would you change it?

Check your knowledge