Lesson 6 - Product Recognition: AI Product Training Station
Challenge 6-1, 6-2, 6-3
Suggested time: 50 minutes. For younger students or first-time users, this lesson may be extended to 60-70 minutes.
Students train a product-image model and use test results to guide additional image collection.
I can explain how examples help train an AI model.
I can train and test a product-image model.
I can choose additional images to improve recognition results.
Engage
2 min.
In the last lesson, the checkout station read a QR code to identify a product. If the QR code is missing and only the product image is visible, can the machine still recognize the product? Let's find out.
Background
2 min.
Image classification uses patterns in an image to predict a category, such as Apple or Banana. The model learns these patterns from labeled example images.
Mind+ V2 provides an Image Classification tool for adding categories, capturing or importing images, training, testing, and exporting a model. Use the prepared project to introduce this workflow.
NOTE: The Mind+ online version does not currently support model training.
Required items:
| Item | Use in this lesson |
|---|---|
| Computer with Mind+ V2 | Runs New Project > Model > Image Classification (M1). Install and open it before class. |
| USB camera | Captures product-card images for training and testing. Confirm permission and camera selection. |
| Four provided product cards | Use the product-image side for Lesson 6. Keep the QR-code side outside every training image. |
| New test images | Prepare new images for each category: Apple, Banana, Candy, and Soda. Keep these images out of the training set. |
| Challenge Cards | Used to record test results and compare predictions before and after retraining. |
Prepare Mind+ and Product Images
3 min.
Preparation tips:
Print the files in product_images. Prepare one new test image for each category.
Complete the classroom setup before students begin Challenge 6-1.
| Before class, check that each group can open Image Classification and use the camera. |
| Product-image check Four product cards ready Image side ready for training One new test image per category Empty-camera test ready |
Open a working image-classification project and demonstrate the results to students. (example-classification-project.mpmodel )
Explore Function
15 min.
Challenge 6-1: Plan How to Teach an AI
| Demonstrate product-image recognition using the prepared project. Ask students to complete the comparison table in Challenge 6-1. |
| Briefly demonstrate the training workflow. Have students use the word bank to complete the process chart in Challenge 6-1. |
| Ask: Which images will you use to teach the AI? Have students complete the four-category data plan, including the images to collect and a new test image for each category. Then discuss their plans. |
As students work through Challenge 6-1, discuss:
• How does recognizing an image differ from reading a QR code?
• How will your image plan help the AI learn the four categories?
NOTE: Use the image side of each card. Keep the initial test images separate from training.
Explore AI Technology
15 min.
Challenge 6-2: Teach the AI About the Products
After the data plan is complete, begin Challenge 6-2.
| Once the plan is ready, ask students to follow the Challenge 6-2 checklist to collect images and train their first model. |
| After training, have students complete Test Using the Training Cards. Have students record the number of training images, the predicted label and confidence for each view, and their observations from the lighting and distance tests. |
| Next, have students complete Test with New Images, including the empty camera view. Ask them to record the results and answer the short reflection prompts before discussion. |
After completing Challenge 6-2, discuss:
• Which test was hardest for the model, and what was different about it?
• What additional examples would you collect to address that problem?
Introduce the key technology
Machine vision uses images to obtain information. Here, image classification predicts which product category an image belongs to.
• Teaching an AI to recognize images is called training. We provide examples with category labels, and the training algorithm uses them to learn patterns that help distinguish the categories.
• The result is a model. When given an image, the model uses the patterns it has learned to predict a product category.
• Training images are examples used to teach the model. To check its response to new examples, use images that were not included in training.
• Confidence indicates how strongly the model favors a prediction; it does not guarantee a correct answer.
Challenge
10 min.
Begin after students have identified a problem in Challenge 6-2.
Challenge 6-3: Improve the Product Model
| Ask groups to choose one problem from their test records. Use Challenge 6-3 to guide additional image collection and retraining. |
| Have students repeat the same test, compare results, and complete the model-file and category-order fields before exporting. |
After completing Challenge 6-3, discuss:
• What examples did you add, and why?
• What changed when you repeated the test?
Summary - Key AI Ideas Review
3 min.
Summarize with students:
• To train AI to recognize products:
Choose categories→Collect and label images→Train a model→Test its predictions.
• When collecting images:
• Include varied examples with different angles, lighting, distances, and backgrounds.
• Keep new test images separate from training. If a test image is added to training, use another new image for testing.
![[Spanish ver.] HuskyLens Cámara Inteligencia Artificial](https://dfimg.dfrobot.com/nobody/makelog/902f49d1d2eb94077e1ba7d8b652cee6_224x164.jpg)
