Module 01 Activity

Scenario

You have been given a folder of images and asked to build something with them. Before any of that, establish what they are and whether you may use them.

What you build

A dataset card written before any modelling, covering permission, incidental content and stated prohibitions.

Steps

1. Load five images and print shape, dtype and channel order for each. Confirm which order your loader produces.
2. Convert one image between RGB and BGR and view both. Record what changed.
3. Answer the permission question for the source, in writing, with the evidence.
4. Inspect EXIF on five files and report what location and timestamp data is present. Strip it.
5. List the incidental content - faces, plates, screens, badges - and decide about each before collecting more.
6. Write the dataset card including may_NOT_be_used_for and known_gaps.

Evidence to hand in

• Shape, dtype and channel order for five images.
• The RGB/BGR comparison.
• The permission answer with evidence.
• The EXIF audit and what was stripped.
• The dataset card.

Review checklist

• The channel order your loader produces is stated, not assumed.
• Permission is answered with evidence, not asserted.
• EXIF was actually stripped, not just noted.
• may_NOT_be_used_for is populated and specific.
• known_gaps names conditions absent from the data.

Mark complete