Metadata-Version: 2.4
Name: smesh
Version: 0.0.1
Summary: Smesh – Mesh→Point-Cloud Sensor Simulator (core skeleton)
Author-email: Smesh Authors <example@example.com>
License: Apache-2.0
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.23
Requires-Dist: laspy>=2.4
Requires-Dist: PyYAML>=6.0
Requires-Dist: pydantic>=2.7
Requires-Dist: typer>=0.12
Requires-Dist: matplotlib>=3.9
Provides-Extra: vtk
Requires-Dist: vtk>=9.2; extra == "vtk"
Provides-Extra: embree
Requires-Dist: trimesh[ray]>=4.4; extra == "embree"
Provides-Extra: laz
Requires-Dist: lazrs>=0.5; extra == "laz"
Dynamic: license-file

# Smesh – filthy-fast mesh→point cloud simulator

This is a minimal, type-annotated implementation of the **critical core components**
for the Smesh simulator:

- `MeshScene` (mesh IO + attribute probing via VTK)
- `Intersector` (VTK OBBTree + optional Embree via `trimesh[ray]`)
- `PointBatch` container
- `AttributeComputer` plug-ins (range, incidence, scan angle, returns, intensity, gps time, color/normal probe, beam footprint)
- `LasWriter` with **streaming** support and dynamic ExtraBytes
- `Sampler` orchestrator which glues everything together

## Installation

Examples below use [`uv`](https://github.com/astral-sh/uv) for reproducible runs, but a standard Python workflow works too:

```bash
python -m venv .venv
source .venv/bin/activate
pip install -e .
```

### Installing straight from GitHub

To pull the latest main branch directly into a virtualenv:

```bash
pip install git+https://github.com/Chiark-Collective/smesh.git
```

For a reproducible install, pin to a tag (e.g., `git+https://github.com/Chiark-Collective/smesh.git@v0.1.0`).

## Quick taste

```python
import numpy as np
from smesh import MeshScene, Sampler, SamplerConfig, AutoIntersector, LasWriter
from smesh.core.intersector import RayBundle

scene = MeshScene(mesh_path="scene.ply")
# Make a tiny fan of rays from above
M = 1000
origins = np.tile(np.array([[0,0,10.0]]), (M,1))
dirs = np.tile(np.array([[0,0,-1.0]]), (M,1))
bundle = RayBundle(origins=origins, directions=dirs, max_range=100.0, multi_hit=False, meta={"gps_time": np.arange(M)*1e-3})

sampler = Sampler(scene, intersector=AutoIntersector(), cfg=SamplerConfig())
writer = LasWriter("out.las", compress=False)
stats = sampler.run_to_writer(writer, ray_batches=[bundle])
print(stats)
```

> Note: for VTK-backed attribute probing and fast VTK intersector, install `vtk`.
> For Embree, install `trimesh[ray]` which pulls `pyembree` (and requires Intel Embree runtime).

## CLI & Preview Runs

Quickly sample a scenario from YAML:

```bash
uv run smesh sample examples/configs/preview/aerial_lidar_preview.yaml
```

Override outputs or attributes on the fly:

```bash
uv run smesh sample examples/configs/preview/mobile_lidar_preview.yaml \
  --output scratch/mobile.las --attribute range --attribute scan_angle
```

Generate fresh preview point clouds and web-friendly images for the key capture modes:

```bash
uv run python scripts/render_examples.py            # preview configs (fast)
uv run python scripts/render_examples.py --full     # full-resolution configs
# or, without uv:
python -m smesh.cli sample examples/configs/preview/aerial_lidar_preview.yaml
```

The script writes point clouds into `examples/outputs/` and PNG composites into `examples/images/`.

## Workflow Examples

| Capture | Preview | Command |
| --- | --- | --- |
| Aerial LiDAR | ![Aerial LiDAR](examples/images/aerial_lidar.png) | `uv run smesh sample examples/configs/preview/aerial_lidar_preview.yaml` |
| Mobile LiDAR | ![Mobile LiDAR](examples/images/mobile_lidar.png) | `uv run smesh sample examples/configs/preview/mobile_lidar_preview.yaml` |
| Total Station | ![Total Station](examples/images/total_station.png) | `uv run smesh sample examples/configs/preview/total_station_preview.yaml` |
| Photogrammetry | ![Photogrammetry](examples/images/photogrammetry.png) | `uv run smesh sample examples/configs/preview/photogrammetry_preview.yaml` |

Each preview config runs in under a minute on CPU-only machines, producing lightweight outputs for quick inspection. Higher fidelity counterparts live in `examples/configs/` and can be regenerated by rerunning `scripts/render_examples.py --full`.

## Output writers

- `LasWriter` streams directly to disk batch-by-batch (compression optional).
- `PlyWriter` and `NpzWriter` buffer incoming batches in memory and emit a single file when `close()` is called. Call `close()` explicitly if you manage the writer yourself.

## Attribute pipeline notes

The sampler now orders attribute computers based on their declared dependencies. Custom attribute chains still work—dependencies like `range_m` for intensity will be inserted automatically so long as a matching producer exists.
