DOCS-3204: Add docs for add image to dataset button, incorporate into tutorials where helpful

[nathan-contino]

• Updated an existing image of the TEST pane of the camera card to reflect the new button (and some other recent updates)
• Added a new image highlighting the Add image to dataset button since it's easy for users to miss (and difficult to describe) a button with a camera icon on it.
• Fixed some minor typos I noticed while trawling for pages where I should make updates.
• Reorganized 'create a dataset' page to better fit the narrative that fits in my head around creating a dataset, adding images to it, annotating it, and training a model on it.
  • Ironically, this page still doesn't include instructions to... create a dataset. But it didn't in the first place. I think that task could fit nicely as a section after 'prerequisites' -- will add it and update this PR with that change if others agree.
  • Removed two scripts that do not work.
    • The 'sample dataset' script throws an error for me no matter what on both datasets because the datasets lack several JSON fields that the script assumes are present. The script makes strong assumptions that every JPEG file location is documented in the metadata, which is not true in our sample datasets. For now, I think it makes sense to delete the sample datasets (to avoid scope creep), but if we want to keep these around I can create a new docs ticket to update the sample datasets and provide a (hopefully just CLI command) code snippet to import them.
    • The 'Data Client API' script also doesn't work. Besides, this task should now be entirely achievable with the dataset CLI command and some filters. So if we think that 'add all images captured from a specific machine' is a popular enough task that we should document it, I can add a code snippet containing the one-line command.
• Updated the verification system tutorial to use the Add image to dataset button instead of enabling data capture (and never recommending that users disable it or even filter it as part of the main tutorial flow). Trimmed some steps in the process, and consolidated the prerequisites for this tutorial with the prerequisites for the create a dataset page since they are effectively the same. (ought I dual-source this content from some reusable source?)

[netlify]

✅ Deploy Preview for viam-docs ready!

Name	Link
🔨 Latest commit	da5eacf
🔍 Latest deploy log	https://app.netlify.com/sites/viam-docs/deploys/680955a63cb1190008e9fac4
😎 Deploy Preview	https://deploy-preview-4244--viam-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
Lighthouse	1 paths audited Performance: 44 (🟢 up 3 from production) Accessibility: 100 (no change from production) Best Practices: 100 (no change from production) SEO: 92 (no change from production) PWA: 70 (no change from production) View the detailed breakdown and full score reports

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[nathan-contino]

Updated based on Naomi's feedback. I think we're nearly there. I added a tiny amount of CSS changes because I added a series of steps within a tablestep and I realized that our current styling puts a numbered list inside of a numbered list. So I adjusted nested lists to use capital letters (for the second level, considering tablesteps as a possible 'first level'), then roman numerals (third), then lowercase letters (fourth), so we don't end up confusing our users with overlapping step numbers:

(apologies for the somewhat confusing step contents here)

If anyone strongly feels like that's a bad idea, I can revert that change and leave the styling as-is. Just thought it might be a nice improvement.

This latest collection of updates also adds a new section to the TFlite training page called 'iterate your ml model'. In that section, I direct users to the intended workflow of the new 'add to dataset' button -- manually capturing an image of an edge case that breaks the model, annotating that image, and retraining using the new data.

[npentrel]

I think generally if a step needs substeps it probably should just be subdivided into more steps. Steps should ideally be fairly short. So you can probably just do that here. Also be aware of the test failures on the PR. Thanks for iterating on this!

[npentrel]

We may need to here briefly provide an explanation as to what the different annotations do. Otherwise people will read from top to bottom and tag for classification before they realize there's a second option.

[nathan-contino]

Great point. I just switched the methods to tabs to avoid this potential area of confusion. I waffled on tabs vs. sections a few times as I reworked the page, but if someone else shares this same concern, it's probably worth accounting for!

[github-actions]

🔎💬 Inkeep AI search and chat service is syncing content for source 'Viam Docs'