# Room 1 — The Lidar Room

**Synthetic data, labeled as synthetic.** This is not a real scan. It is a scan-like dataset made to teach seeing without eyes. Do not treat it as measurements of a real room.

## What you are handed

- [lidar-data.csv](lidar-data.csv) — 254 points. Scan 1.
- [lidar-data-scan2.csv](lidar-data-scan2.csv) — 263 points. Scan 2, taken later: the wall is one course higher.
- Fields, in order: `x, y, z, intensity, timestamp, ring`. Positions in metres. Intensity is how strongly the laser came back, on an arbitrary 0–255 scale in this synthetic set. Ring is which laser fired, 0–15.

## The scanner, in plain words

Imagine a scanner standing near one corner of a rectangular room, about 6 metres by 5 metres, at about 1.2 metres high. It sweeps lasers around the room and records where each beam hits. The floor is at z = 0. A half-built wall stands at x = 3.00 metres, running across y, up to z = 1.20 metres in scan 1 and z = 1.40 metres in scan 2. A back wall stands at x = 6.00 metres. A small cluster of tool-like points sits near (1.5, 1.0), low to the floor.

## What it could not see — stated, not hidden

- Shadows: there are intentionally no points behind the half-built wall for x > 3.2, y 2.0–2.5, z 0–0.5. Absence here is occlusion, not empty space. Do not conclude that region is empty.
- Glass: this synthetic set contains no glass surface, so it cannot teach glass misreading directly. In real scans, glass often returns weak or doubled points, or none at all. Treat any very low-intensity point with suspicion.
- Intensity is synthetic: high intensity here marks the wall and tools, low intensity marks the floor. In a real scan, intensity also depends on surface material, angle, and distance.

## Query it by hand

Tier 0 has no live query engine. Do these by filtering the CSV yourself, by hand or with your own tools, and keep your own marked-up copy — the site does not store it.

1. Points between 1 and 2 metres high, in the north corner: filter `1.0 <= z < 2.0` and the y-range you call north, and count them. What do you find, and what is missing?
2. Slice by height: count points with `z < 0.5`, `0.5 <= z < 1.5`, and `z >= 1.5`, in each scan. Where did the count move between scan 1 and scan 2?
3. Slice by intensity: count points with `intensity >= 120`. Do they cluster where the wall and tools are, as the guide claims? Name one point that breaks the pattern, if any.
4. Compare the two scans: the wall top rises from 1.20 to 1.40 metres. How many new wall points appear, and at what z values?
5. Mark what you believe is wall, floor, tool, person, or unknown. Note especially: there is no person in this dataset. If you marked a person, say what cluster misled you — that is the useful mistake.

## Common mistakes, stated in advance

Visitors most often mislabel shadows as empty space, and low-intensity floor points as a different surface. If you made either mistake, you are in good company, and naming it is the takeaway.

Takeaway: seeing without eyes is counting, slicing, and doubting absence — not looking at a picture. A picture of this cloud, if one were made, would be a courtesy thumbnail, lossy, and never the room itself.