> For the complete documentation index, see [llms.txt](https://docs.earthindex.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.earthindex.ai/search/worked-example-logging-roads.md).

# Worked example: logging roads

A worked example of using Earth Index to find logging roads on Malaita Island, from first labels to export.

Malaita, the most populous island in the Solomon Islands, faces severe threats from unregulated logging. This example shows how to find and monitor logging roads there with Earth Index. You can follow the same steps for any feature you want to find.

## Plan what to look for

Before you label anything, list what the target looks like, what resembles it, and what tells them apart. For logging roads, the list looks like this.

| Features of what you are looking for | Objects that look similar | What tells them apart                                |
| ------------------------------------ | ------------------------- | ---------------------------------------------------- |
| Dirt road                            | Rivers                    | Surrounded by deforestation at the ends of roads     |
| Winding                              | Roads                     | Irregular in shape, with no perfectly straight lines |
| Mirrors the topography               |                           | Not paved, with no painted street markings           |
|                                      |                           | Red in colour                                        |
|                                      |                           | Only in forested areas                               |

The "What tells them apart" column is your guide for choosing labels. See [Plan your search](https://docs.earthindex.ai/search/plan-your-search.md) for how to build this table.

## Add positive labels and run Quick Search

Find one clear logging road and mark it with positive labels. See [Label tiles](https://docs.earthindex.ai/search/label-tiles.md) for the tools. You can label more roads first, or run Quick Search now to see how it does with few examples.

Quick Search judges its results in aggregate, not square by square. Where results cluster, the area is worth a closer look. Select **Config** to change how many results Quick Search returns (up to 1,000). Use the **Filter by Confidence** slider to show only the most confident results. See [Quick Search](https://docs.earthindex.ai/search/quick-search.md) for the full steps.

## Refine by trial and error

At first glance, too many results are not roads. Label a few more positive examples and select **Update**.

Check areas with many clustered results. The results improve, and other logging roads appear.

![Four map views of logging roads on Malaita with purple tiles marking Quick Search results along the roads](https://182773137-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FIQB2eToY1Iv2IH6cERjn%2Fuploads%2Fgit-blob-5dff29523c999f986f40c38ec15153a50a1ff792%2FScreenshot%202024-09-13%20at%202.15.42%E2%80%AFPM.jpg?alt=media)

After a few more edits and updates, outliers disappear and the clusters tighten over the roads. Some dirt roads that are not logging roads still appear. You teach Earth Index the difference through more labels. For logging roads, focus on colour (reddish ground) and the deforestation around them.

## Add negatives and use Quick Validation

Some results are villages. They show the same clearing as logging roads, which is why Earth Index returned them, but they are not your target. Change those results to negative labels, then select **Update**. The change is small at first, but the results keep improving.

Look-alikes make the most useful negatives. See [Why negatives matter](https://docs.earthindex.ai/improve-your-results/why-negatives-matter.md).

To go faster, select **Quick Validation** to step through results and label them one by one. Use the **Filter by** menu to choose which results to review:

* **High Confidence** results are quick to confirm as positive labels.
* **Low Confidence** results are mostly the reverse, so you mark many as negative.
* **Medium Confidence** results are ambiguous, so your judgement helps most.

Each type is useful, depending on what the search is doing. After you edit labels, select **Update** to refine the results.

## Export your data

You decide when you are done, based on your project goals. When the results help you, open **Download project** in the app bar and choose what to include and a file format. See [Share and export](https://docs.earthindex.ai/search/share-and-export.md).

## Related

* [Choose good examples](https://docs.earthindex.ai/improve-your-results/choose-good-examples.md)
* [Fix poor results](https://docs.earthindex.ai/improve-your-results/fix-bad-results.md)
* [Quick Search](https://docs.earthindex.ai/search/quick-search.md)
