Disparity to Ordered Point Cloud Converter

Convert TIFF disparity images to ordered point clouds.

Installation

pip install -e examples/python/disparity_to_ordered_point_cloud

Usage

cd examples/python/disparity_to_ordered_point_cloud/disparity_to_ordered_point_cloud
python3 disparity_to_ordered_point_cloud.py <disparity_dir> <details_dir> <output_dir> [--split]

Parameters

  • disparity_dir: Path to directory containing TIFF disparity images
  • details_dir: Path to directory containing YAML details files
  • output_dir: Directory where ordered point clouds will be saved
  • --split: Optional flag to save XYZ dimensions as separate text files

Examples

# Generate 3-channel TIFF point clouds (default)
python3 disparity_to_ordered_point_cloud.py /path/to/disparity /path/to/details /path/to/output

# Generate separate text files for each XYZ dimension
python3 disparity_to_ordered_point_cloud.py /path/to/disparity /path/to/details /path/to/output --split

Output

  • Format: 3-channel TIFF files (default) or separate text files (with --split)
  • Location: Specified output directory
  • Naming: Maintains original disparity image naming convention

Features

  • Convert TIFF disparity images to ordered point clouds
  • Support for both unified and split output formats
  • Preserves spatial ordering of 3D points
  • Compatible with various 3D visualization tools

Output Formats

Default (3-channel TIFF)

  • Single TIFF file per disparity image
  • Three channels representing X, Y, Z coordinates
  • Maintains pixel-to-point correspondence

Split Format (--split option)

  • Three separate text files per disparity image
  • Individual files for X, Y, Z dimensions
  • Human-readable format for analysis

Requirements

  • TIFF disparity images from depth_to_disparity converter
  • YAML details files from Hammerhead data collection
  • Sufficient storage space for point cloud data

Troubleshooting

  • Missing details files: Ensure details directory contains corresponding YAML files
  • File overwrite warning: Existing files in output directory will be overwritten
  • Large output files: Point clouds can be very large - ensure adequate storage
  • Memory issues: Processing large disparity images may require significant RAM

Press Ctrl+C to stop conversion.