Module 4: Convolutional neural networks for vision

Module 4: Convolutional neural networks for vision#

AINS6003 — Deep Learning & Neural Networks

Guided study deck (about 90 minutes)

Essential question: Why do convolutions, pooling, and translation equivariance matter for image understanding?

Why This Matters#

A computer-vision team is deciding whether a small CNN is adequate for grayscale inspection images before moving to larger pretrained models

You do not need a computer science background to use this module. Focus on reading the AI workflow, asking precise questions, and explaining what the evidence does and does not support.

Guided Study Path#

Time	Segment	Your purpose
0-10	Orientation and stakes	Connect the topic to a professional decision.
10-25	Conceptual model	Build intuition before code or formulas.
25-40	Worked example	Translate vocabulary into a small concrete case.
40-55	Evidence and interpretation	Read outputs, metrics, or artifacts carefully.
55-70	Guided practice	Change one variable and observe the result.
70-82	Risk, limits, and communication	Name what could go wrong and explain it clearly.
82-90	Assignment planning	Confirm the deliverable, rubric, and next step.

Learning Outcomes#

explain the core technical idea in precise neural-network vocabulary
connect architecture and training choices to the shape of the data and task
run or interpret the module lab as reproducible evidence
identify limitations, failure modes, and next steps for a defensible experiment

By the end, you should be able to explain the idea without hiding behind jargon and identify what evidence would make a recommendation stronger.

Plain-Language Framing#

Complete this sentence before introducing the technical terms:

This method helps a professional decide whether ______ because it uses ______ as evidence.

Revisit the sentence at the end of your study session and improve it with precise module vocabulary.

Core Vocabulary#

local receptive fields: define it in one sentence, then connect it to the scenario.
weight sharing: define it in one sentence, then connect it to the scenario.
feature maps and channels: define it in one sentence, then connect it to the scenario.
padding, stride, pooling, and flattening: define it in one sentence, then connect it to the scenario.
classification heads for image features: define it in one sentence, then connect it to the scenario.

🧑‍🌾 SAMWISE — Student note

Pause after each term and write your own example before continuing. This is prewritten guidance; no reply is expected.

Conceptual Model#

Use a three-part model:

Input: What information is available?
Transformation: What does the AI or analytic method do to the information?
Decision: What human or organizational action could change because of the result?

Use this model even if Python is new to you; it separates professional reasoning from code syntax.

Worked Example Setup#

Apply the professional scenario as your example case. Ask yourself:

Who owns the decision?
What evidence would they trust?
What would count as a bad recommendation?
What would a cautious first experiment look like?

Record concise answers before opening the lab.

Method Pattern#

Translate image dimensions into tensor shapes before designing the classifier head.
Use convolution and pooling choices that match expected object scale.
Inspect intermediate shapes and, when possible, activations.
Separate image preprocessing assumptions from model architecture assumptions.

Practice this repeatable professional move throughout the program.

Lab Bridge#

Lab notebook: Module 4 Lab: CNN feature maps

Open the lab from your private course repository in Codespaces or Colab. Run all cells first, then change exactly one value, threshold, feature, or assumption. Focus on observation and interpretation rather than writing code from scratch.

Reading Lab Outputs#

When you see a number, plot, table, or printed result, ask four questions:

What changed?
Is the change large enough to matter?
What assumption produced the result?
What would be needed before using this outside the toy setting?

Guided Practice#

Work independently or compare observations with a study partner:

Run the lab unchanged.
Change one small input or parameter.
Capture the before/after result.
Write a two-sentence interpretation for a nontechnical stakeholder.

Save both the evidence and your interpretation in your private course repository.

Common Failure Modes#

Flattening too early and discarding spatial structure.
Using augmentation that changes the label meaning.
Overlooking class imbalance, acquisition artifacts, or scanner/source shift.
Claiming visual understanding from a toy shape trace alone.

Name these risks explicitly in your notes before trusting the output.

Reflection Checkpoint#

Pause around the 60-minute mark and answer:

What did the method make easier to see?
What did the method hide or simplify?
Who might be harmed by a confident but wrong interpretation?
What evidence would make you more comfortable recommending action?

Assignment Preparation#

Module 4 Assignment: CNN architecture memo

Design a CNN for a small grayscale image classification problem
Specify convolution, activation, pooling, flattening, and output-head choices
Run the starter model and inspect output shapes after each stage
Explain how local structure and weight sharing support the design

Before beginning the assignment, identify the artifact you will produce, the evidence you must include, and the limitation you must state.

Rubric Self-Check#

Use these plain-language checks before submitting:

Correct: terms and results are used accurately.
Evidence-based: claims point to notebook output, scenario facts, or documented assumptions.
Context-aware: the recommendation fits the stakeholder decision.
Honest: limitations and risks are named clearly.

Closing Reflection#

Write one paragraph:

Explain what this module helps you decide, what evidence the lab produced, and what you would still need before trusting the result in a real organization.

Save the paragraph as the opening of your assignment memo or as a study note for revision.

Choose Your Study Path#