---
title: "How image models work"
lesson_id: "03"
---

# How image models work

- **Lesson ID:** 03
- **Goal:** Distinguish image models that inspect pixels from models that generate pixels.
- **Human lesson:** [03-image-models.html](03-image-models.html)

## Prerequisites

- Lesson 01, or a basic understanding of model inputs and outputs.
- Familiarity with a digital photo and a label such as shoe.

## Inputs, outputs, and artifacts

- **Inputs:** Pixel grids or image patches, plus an image request for generative work.
- **Outputs:** Labels, locations, masks, captions, answers, or new pixel grids.
- **Artifacts:** Progressive `.flow` diagrams for grid input, seeing, and making.

## Agent build steps

1. Describe an image as a grid of visual values or patches.
2. Explain the see job through classification, detection, and segmentation.
3. Place Fashion-MNIST ClothingCNN from [the forward course](../pytorch-core-ai-xcode/) in the see-and-classify group.
4. Contrast it with a generative image model that outputs new pixels.
5. State that larger vision systems can change the output while retaining an image-shaped input path.
6. Link [From the metal to the model](../core-ai-from-metal-back/) to connect model choices to runtime and silicon profiling.

## Constraints

Keep the image explanation practical and light. Avoid vision papers, math deep-dives, CUDA, and quizzes. Keep both prior-course links relative. Do not use em dashes. Use HTML `.flow` diagrams, not Mermaid.

## Key concepts

- A digital image is a grid of visual values. Some systems also use patches.
- See jobs turn pixels into information: a label, boxes, or a mask.
- Fashion-MNIST ClothingCNN is see-and-classify. It does not draw a new shirt.
- Make jobs take a prompt or edit and produce new or changed pixels.

## Takeaways

- You can split image models into see versus make.
- You can place ClothingCNN in the classify bucket, not the generative one.
- You can name classifier, detector, and segmenter outputs without mixing them.

## Acceptance checks

- Pixel grid or patch input is clear.
- See and make are separated with concrete app jobs.
- ClothingCNN is explicitly a Fashion-MNIST image classifier, not a generator.
- Both prior courses are linked and the Agent brief link is present.
- [ ] Proceed to [lesson 04 brief](04-how-they-differ.llms.md).