Can You Design An AI Model For Cupcake And Cookie Identification?

Objective:

Students will create an AI model to identify and classify cupcakes and cookies, selecting from several AI tools to explore machine learning. Using key tools to train their AI model, they’ll compare traditional classification methods with modern AI technology. A big thanks to SugarLove Bakery for providing the images that make this challenge even sweeter!

Materials List:

• Tools - Choose One:

• Nous AI Set (with camera and sensors)
• Google’s Teachable Machine (free web-based platform- you will need access to a camera)
• Cupcake and cookie samples (real or images - recommend at least 30 of each) 
• Laptop or tablet
• Internet connection for research and image datasets
• Camera that connects to your device (only if using Google’s Teachable Machine)
• Printed or digital resources on cupcake and cookie characteristics (some information is available in the background knowledge to guide you but it is encouraged to have students gather and compile their own list.)
• Student Worksheet: Can You Design An AI Model

Introduction:

In this activity, students will learn how AI can assist in identifying and classifying different objects, focusing on cupcakes and cookies. By building and training their own AI models, students will explore machine learning principles and apply them to a real-world context. This challenge connects to NGSS standards and helps students develop skills in data analysis, computational thinking, and classification systems.

Background Knowledge:

Before starting the challenge, students need to understand how objects can be classified based on characteristics like shape, texture, and ingredients. Watching videos or reading materials that explain the differences between cupcakes and cookies, such as their physical features or ingredients, will help them select identifiable traits for their AI models. This foundational knowledge ensures that students choose accurate features for training their AI tools.

Additionally, students should learn how classification systems work, whether traditional or AI-based. While traditional classification methods use specific rules and observable characteristics (like a dichotomous key), AI models rely on pattern recognition from data. This knowledge will help students compare and contrast different identification methods.

Information on Cupcake vs Cookie characteristics:

When designing your AI model for classifying cupcakes and cookies, it’s essential to understand their distinct characteristics. Cupcakes are individual-sized cakes with a soft, moist texture and are typically topped with frosting or decorations, making them ideal for special events and presentations. They are visually appealing with their dome shape and often come with intricate toppings like sprinkles or fondant.

Cookies, on the other hand, tend to have a crispy exterior and a chewy or crumbly interior. They come in a variety of flavors, such as chocolate chip, oatmeal, and peanut butter, and their simplicity makes them a favorite comfort food. 

Here is a list of distinguishing characteristics that an AI model can use to identify and classify cupcakes and cookies:

Cupcake Characteristics:

1. Shape & Structure:
• Dome-shaped with a muffin-like base.
• Individual portions baked in wrappers or paper cups.
2. Texture:
• Soft, light, and fluffy inside.
• Typically has a moist crumb structure.
3. Frosting/Decoration:
• Topped with frosting (buttercream, ganache, etc.).
• May include decorations like sprinkles, edible flowers, or fondant.
4. Use Case:
• Served at special events (e.g., birthdays, weddings) due to elegant presentation.
5. Color Variability:
• Colorful frosting and toppings.

Cookie Characteristics:

1. Shape & Structure:
• Generally flat, round, or irregular in shape.
• Comes in various sizes (from bite-sized to large cookies).
2. Texture:
• Can be crispy on the outside with a chewy interior, or entirely crunchy.
3. Presentation:
• Lacks frosting (though sometimes has a light glaze or drizzle).
• Less visually elaborate than cupcakes.

References for additional information: Flavor Insider and DeliFo

These characteristics offer key visual and textural differences that your AI model can leverage. For example, the presence of frosting or a rounded shape could signal a cupcake, while a flat shape with uneven texture might indicate a cookie. A well-trained AI model can classify these baked goods by detecting these features from images, helping it distinguish between the two types.

These differences will be critical when selecting features for your AI model. For example, cupcakes might be identified based on their frosting color, shape, or decorations, while cookies could be classified by texture, size, or patterns like chocolate chunks. Understanding these attributes will allow your AI system to effectively learn the patterns required for accurate identification.

Incorporating these distinctions into your model will also help students appreciate how AI classification differs from traditional methods as they learn to train their AI on varied features such as shape, texture, or color to achieve better accuracy in the identification task. 

Design:

1. Cupcake and Cookie Characteristics:
• Review the visual and textural differences between cupcakes and cookies (e.g., frosting, size, ingredients).
• Introduce a traditional method of classification, such as a checklist or dichotomous key, to help categorize baked goods based on their attributes.
2. AI and Machine Learning Introduction:
• Explain how AI systems can "learn" to recognize patterns in objects like cupcakes and cookies through machine learning.
• Provide real-world examples of AI applications in food, such as smart kitchen tools or food delivery apps.
3. Tool Selection:
• Nous AI: Students can feed cupcake and cookie images into Nous AI, using the camera, and train it to identify them based on visual characteristics like shape and toppings.
• Teachable Machine: Upload images of cupcakes and cookies to Google's Teachable Machine, which will use image recognition to learn the differences.
4. Training the AI Model:
• Students will collect or download images of cupcakes and cookies, ensuring they have a variety of examples to train their model.
• Depending on the selected tool, they will input these images or data and label them to help the AI recognize patterns like color, size, or icing.
5. Testing & Evaluation:
• Once the AI model is trained, students will test its accuracy by inputting new images of cupcakes and cookies to see how well the AI identifies them.
• Compare the AI’s performance with traditional methods, discussing the accuracy and reliability of each approach.

Presentation and Reflection:

1. Presentation:
• Students will present their AI models, explaining the tool they selected, how they trained the model and the results of their testing.
• They will compare their AI’s ability to classify cupcakes and cookies to traditional methods like using a checklist or dichotomous key.
2. Reflection:
• Reflect on the learning process and discuss the advantages and disadvantages of using AI for object classification.
• Consider ethical questions such as: How might AI change how we classify or judge food? What are the implications for jobs in the food industry?
• Did you uncover any bias in your AI model?  (An example might be the inability to identify a cupcake that has a piece of cookie on top of it.)

Optional Extension Activities:

• Expand the Dataset: Collect more diverse cupcake and cookie samples, including variations in decoration, size, and ingredients, to improve the AI’s accuracy.
• Tool Comparison: Try a different tool (e.g., switch from Nous AI to Teachable Machine) and compare the outcomes.
• Real-World Connection: Explore how AI could be used in the food industry for tasks like sorting products in a bakery or developing new recipes.