point-tiler

KO | EN

gitlite — search

#python #react #ai #go #golang #machine-learning #rust #docker #windows #awesome #cli #pytorch #openai #open-source

(★ 79)

A tool for converting PointCloud data (las/laz and csv) into "3D Tiles v1.1".

point-tiler Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Point Tiler

A tool for converting point cloud data (LAS/LAZ/CSV) into 3D Tiles v1.1.

**English**|[日本語](./README.ja.md)

![alt text](https://raw.githubusercontent.com/MIERUNE/point-tiler/main/images/README_image.png)

## Point Tiler?

Point Tiler stands out from existing open-source point cloud tilers in the following ways:

- **Native 3D Tiles 1.1 output** — Directly outputs GLB (glTF Binary) files as specified in 3D Tiles 1.1. Most alternatives still rely on the legacy `.pnts` format (3D Tiles 1.0) or offer only experimental 1.1 support.
- **Versatile compression** — Supports quantization (`KHR_mesh_quantization`), meshopt compression (`EXT_meshopt_compression`), and GZIP to significantly reduce output file sizes.
- **LAZ support** — Reads compressed LAZ files with parallel decoding, which is essential for real-world workflows where LAZ is the dominant format.
- **Large-scale data handling via external sort** — Automatically switches between in-memory and external sort workflows based on a conservative in-memory estimate, so you don't need to manually split files to fit in RAM.
- **Fast conversion** — Built in Rust with Rayon-based parallelism, delivering high throughput for large datasets.

## Features

- Converts LAS/LAZ/CSV point cloud data to 3D Tiles v1.1 format
- Built-in PROJ library supporting a wide range of coordinate systems
- High-performance parallel processing with Rayon
- Streaming processing for large-scale data
- Voxel-based point decimation for efficient LOD
- GZIP compression support for output tiles
- Quantization support for smaller file sizes (`KHR_mesh_quantization`)
- Meshopt compression for optimized GPU delivery (`EXT_meshopt_compression`)

## Project Structure

```
point-tiler/
├── app/                       # Main CLI application
├── pcd-core/                  # Core point cloud data structures
├── pcd-parser/                # LAS/LAZ/CSV file parsers
├── pcd-exporter/              # Tiling and 3D Tiles export
├── coordinate-transformer/    # Coordinate transformation (PROJ-based)
└── cesiumtiles-gltf/          # glTF/GLB I/O for 3D Tiles output
```

## Install

Enter the following command.

```sh
curl -sSf https://raw.githubusercontent.com/MIERUNE/point-tiler/main/install.sh | bash
```

Currently, it is not available on Windows.
However, if Rust is installed and Cargo is available, it is possible to run it using the `cargo` command.

### For Developer

Rust must be installed. You can easily install it from the following page.

[Getting started](https://www.rust-lang.org/learn/get-started)

After installing Rust, download this repository.

## Usage

### Options

| Option                 | Description                                                                                                                                                |
| ---------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `--input`, `-i`        | Input file path(s). Supports `.las`, `.laz`, `.csv`, `.txt`. Multiple files can be specified.                                                              |
| `--output`, `-o`       | Output folder path. Outputs `tileset.json` and GLB files.                                                                                                  |
| `--input-epsg`         | EPSG code of the input coordinate system. Supports any EPSG code via PROJ.                                                                                 |
| `--output-epsg`        | EPSG code of the output coordinate system. Typically EPSG:4979 (WGS84 Geographic 3D) for Cesium.                                                           |
| `--min`                | Minimum zoom level (default: 15)                                                                                                                           |
| `--max`                | Maximum zoom level (default: 18)                                                                                                                           |
| `--max-memory-mb`      | Memory budget in MB used for workflow selection. In-memory mode is chosen only when `estimated processing size × 5` fits within this value (default: 4096) |
| `--threads`            | Number of threads for parallel processing (default: number of CPU cores)                                                                                   |
| `--quantize`           | Enable quantization for smaller GLB files (`KHR_mesh_quantization`)                                                                                        |
| `--meshopt`            | Enable meshopt compression (`EXT_meshopt_compression`)                                                                                                     |
| `--gzip-compress`      | Enable GZIP compression for output tiles                                                                                                                   |
| `--disable-decimation` | Disable decimation during intermediate file generation and keep original point density                                                                     |

### Example

```sh
ptiler --input app/examples/data/sample.las \
    --output app/examples/data/output \
    --input-epsg 6677 \
    --output-epsg 4979 \
    --min 15 \
    --max 18 \
    --max-memory-mb 8192 \
    --threads 8 \
    --quantize \
    --meshopt \
    --gzip-compress
```

### Benchmark

| Environment |                          |
| ----------- | ------------------------ |
| Machine     | Apple M1 Max / 64 GB RAM |
| OS          | macOS (Darwin 23.4.0)    |

**Command:**

```sh
ptiler --input /path/to/data/*.las \
    --output /path/to/output \
    --input-epsg 6677 \
    --output-epsg 4979 \
    --min 15 \
    --max 18 \
    --max-memory-mb 8192 \
    --threads 8 \
    --quantize \
    --meshopt \
    --gzip-compress
```

| Input      |              |
| ---------- | ------------ |
| Files      | 41 LAS files |
| Total size | ~8.0 GB      |

| Result       |                              |
| ------------ | ---------------------------- |
| Total time   | **3 min 57 sec** (236.9 sec) |
| Output tiles | 278 tiles                    |

### Memory Workflow Selection

`--max-memory-mb` is not a hard runtime memory cap. It is a conservative workflow-selection threshold.

- `estimated processing size × 5 <= max-memory-mb` selects the in-memory workflow
- Otherwise, point-tiler falls back to the external sort workflow

This headroom accounts for point buffers, grouping structures, and parallel intermediate allocations that can make actual RSS significantly larger than the raw point data size.

During conversion, temporary intermediate files are generated on disk. In current benchmarks, their peak size is typically around 1.2x to 1.5x the input LAS size, and they are deleted automatically after processing finishes.

### Attributes

At present, 3D Tiles output uses only XYZ and RGB. Other point attributes such as intensity, return number, classification, scan angle, point source ID, and GPS time are ignored during GLB generation.

### Coordinate Systems

This tool uses the PROJ library for coordinate transformation, supporting any EPSG code.

**Common input coordinate systems:**
- Japan Plane Rectangular CS (EPSG:6669-6687)
- UTM zones (e.g., EPSG:32654 for UTM zone 54N)
- Any other projected or geographic coordinate system

**Recommended output coordinate systems:**
- EPSG:4979 (WGS84 Geographic 3D) - Standard for Cesium
- EPSG:6697 (JGD2011 Geographic 3D) - For Japan-specific applications

### Zoom Levels

This tool has a unique concept called a `zoom level` which represents a planar area roughly the same size as 2D tiles such as raster tiles or Google Photorealistic 3D Tiles.
The geometric error varies depending on the zoom level for example, it is approximately 64.0 at zoom level 15, about 8.0 at zoom level 18, and about 1.0 at zoom level 21.
The output tiles are thinned out so that one point is stored in a voxel grid of `Geometric Error × 0.1`.

### CSV/TXT Format

It supports .csv and .txt file extensions.

In CSV format, any columns other than XYZRGB will be ignored.
Also, the column names must be "x", "y", "z" and "r", "g", "b" or "red", "green", "blue".
(The case of the letters does not matter.)

For example, the following data is valid.

```csv
"X","Y","Z","Intensity","ReturnNumber","NumberOfReturns","ScanDirectionFlag","EdgeOfFlightLine","Classification","Synthetic","KeyPoint","Withheld","Overlap","ScanAngleRank","UserData","PointSourceId","GpsTime","Red","Green","Blue"
-5599.971,-35106.097,3.602,680.000,1.000,1.000,1.000,0.000,1.000,0.000,0.00 0,0.000,0.000,-4.000,0.000,95.000,188552.207,19532.000,20046.000,20560.000
-5599.976,-35111.458,3.440,715.000,1.000,1.000,1.000,0.000,1.000,0.000,0.00 0,0.000,0.000,-4.000,0.000,95.000,188552.287,20560.000,21074.000,21588.000
-5599.978,-35115.695,3.543,311.000,1.000,1.000,1.000,0.000,1.000,0.000,0.00 0,0.000,0.000,-4.000,0.000,95.000,188552.347,19018.000,19275.000,20303.000
-5599.992,-35119.757,3.603,722.000,1.000,1.000,1.000,0.000,1.000,0.000,0.00 0,0.000,0.000,-4.000,0.000,95.000,188552.407,18504.000,20046.000,20817.000
-5599.992,-35129.327,3.431,505.000,1.000,1.000,1.000,0.000,1.000,0.000,0.00 0,0.000,0.000,-4.000,0.000,95.000,188552.547,18504.000,19789.000,21074.000
```

## Roadmap

- [ ] Automatic CRS detection from input files
- [ ] Public library API for external integration
- [ ] PLY format input

## License

[MIT License](./LICENSE)

## Authors

- Satoru Nishio (@nokonoko1203)
- And all contributors

## Use of Third-Party Source Code

This project incorporates and modifies source code from the following third-party projects:

- [PLATEAU GIS Converter](https://github.com/MIERUNE/plateau-gis-converter)
  - Repository: [PLATEAU GIS Converter](https://github.com/MIERUNE/plateau-gis-converter)
  - Original License: [MIT License](https://github.com/MIERUNE/plateau-gis-converter/blob/main/LICENSE)

### Modifications Made

We have made the following modifications to the source code from [PLATEAU GIS Converter](https://github.com/MIERUNE/plateau-gis-converter):

- [nusamai-gltf](https://github.com/MIERUNE/plateau-gis-converter/tree/main/nusamai-gltf) is included as [cesiumtiles-gltf](./cesiumtiles-gltf/).

The modified code retains its original copyright and is used in accordance with the MIT License. For details, see the LICENSE file.

point-tiler

point-tiler Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

point-tiler

point-tiler Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README