SLAM Pipeline Output Visualizer

A standalone Python script for visualizing SLAM (Simultaneous Localization and Mapping) pipeline outputs using the rerun-sdk. This tool provides an interactive 3D visualization of point clouds, camera trajectories, floor detection, closest point analysis, and video synchronization.

Features

🎯 Core Visualization

• Point Cloud Rendering: Static and temporal point cloud visualization with color support
• Camera Trajectory: 3D camera path with temporal playback
• Floor Detection: Floor plane visualization with grid overlay
• Closest Point Analysis: Real-time proximity warnings with distance visualization
• Video Synchronization: Synchronized video playback with trajectory data

🎨 Advanced Features

• Segmentation Support: Color point clouds by segmentation class IDs
• Warning System: Multi-level proximity alerts (Normal, Caution, Strong, Critical)
• Directional Guidance: Human-readable direction descriptions
• View Cones: Camera field-of-view visualization
• Spatial Subsampling: Memory-efficient point cloud rendering
• Floor Highlighting: Automatic floor point identification and highlighting

📊 Interactive Elements

• Real-time Plots: Distance and warning level charts
• Color-coded Warnings: Visual proximity alerts with appropriate colors
• Temporal Playback: Step-through trajectory with synchronized data
• Separate Segmentation Views: Individual class visualization options

Installation

Prerequisites

• Python 3.7+
• rerun-sdk
• numpy
• plyfile

Install Dependencies

pip install rerun-sdk numpy plyfile

Optional Dependencies

For enhanced point cloud processing:

pip install open3d

Usage

Basic Usage

python visualize_slam_output.py /path/to/slam_analysis_output_dir

Full Usage

python visualize_slam_output.py --color_pointcloud_by_classIds --show_segmentation_masks_separately --show-yolo-detections /path/to/slam_analysis_output_dir

Command Line Options

Visualization Toggles

--no-point-cloud          # Disable point cloud visualization
--no-trajectory          # Disable trajectory visualization
--no-floor              # Disable floor plane visualization
--no-closest-points     # Disable closest points analysis
--no-view-cones         # Disable view cone visualization
--no-trajectory-path    # Disable static trajectory path
--no-distance-lines     # Disable distance lines to closest points
--no-highlight-floor    # Disable floor point highlighting
--no-video              # Disable video visualization

Point Cloud Options

--color_pointcloud_by_classIds     # Enable coloring by class IDs
--show_segmentation_masks_separately  # Show segmentation classes separately
--subsample-percentage 0.5         # Percentage of points to keep (0.0-1.0)
--no-subsampling                   # Disable point cloud subsampling

Warning System

--warning-distance-critical 0.2    # Critical warning threshold (meters)
--warning-distance-strong 0.4      # Strong warning threshold (meters)
--warning-distance-caution 0.6     # Caution warning threshold (meters)
--no-warnings                      # Disable proximity warning system
--no-directional-warnings          # Disable directional warnings

Visualization Parameters

--floor-threshold 0.05             # Floor point identification threshold
--grid-size 2.0                    # Floor grid size (meters)
--cone-angle 45.0                  # View cone angle (degrees)
--cone-length-factor 1.5           # View cone length factor
--max-cone-length 2.0              # Maximum cone length (meters)

Examples

Complete Visualization

python visualize_slam_output.py --color_pointcloud_by_classIds --show_segmentation_masks_separately ./outputs 

Basic Visualization

python visualize_slam_output.py ./enhanced_slam_outputs/slam_analysis_20240115_143045

Minimal Visualization (Point Cloud Only)

python visualize_slam_output.py ./outputs --no-view-cones --no-floor --no-video

Enhanced Point Cloud with Segmentation

python visualize_slam_output.py ./outputs --color_pointcloud_by_classIds --show_segmentation_masks_separately

Custom Warning Thresholds

python visualize_slam_output.py ./outputs --warning-distance-critical 0.15 --warning-distance-caution 0.8

Memory-Optimized for Large Datasets

python visualize_slam_output.py ./outputs --subsample-percentage 0.8

Disable Warning System

python visualize_slam_output.py ./outputs --no-warnings

Data Format Support

Input Files

The script automatically detects and loads the following file types:

Point Clouds

• slam_analysis_pointcloud.ply
• incremental_analysis_detailed_pointcloud.ply
• Intermediate step point clouds: intermediate/step_*/pointcloud.ply

Trajectory Data

• slam_analysis_trajectory.txt
• incremental_analysis_detailed_trajectory.txt

Floor Detection

• slam_analysis_floor_data.json / .csv
• incremental_analysis_detailed_floor_data.json / .csv

Closest Points Analysis

• slam_analysis_closest_points.json / .csv
• incremental_analysis_detailed_closest_points.json / .csv

Video

• Automatically loaded from pipeline configuration
• Supports rerun-compatible MP4 files

Metadata

• metadata.json: Pipeline configuration and file paths

Warning System

The visualizer includes a sophisticated proximity warning system with four levels:

Level	Distance	Color	Description
Normal	>0.6m	Green	Safe distance
Caution	0.4-0.6m	Yellow	Moderate proximity
Strong	0.2-0.4m	Orange	Close proximity
Critical	<0.2m	Red	Immediate danger

Directional Warnings

When enabled, the system provides human-readable direction descriptions:

• "Object 0.3m ahead-right"
• "Object 0.15m to your left"
• "Object 0.25m behind-below"

Performance Optimization

Point Cloud Subsampling

• Percentage-based: Configurable reduction (default: 50%)
• Final step preservation: Always shows 100% of points in final temporal step
• Spatial consistency: Maintains point distribution across space
• Memory management: Automatic subsampling for large datasets

Temporal Data Handling

• Step mapping: Intelligent mapping between trajectory poses and point cloud steps
• Video synchronization: Step-based frame mapping for smooth playback
• Memory efficiency: Temporal point clouds only show current frame

Troubleshooting

Common Issues

Rerun Connection Failed

# Start rerun viewer manually
rerun --serve-web
# Then run the script
python visualize_slam_output.py ./outputs

Video Not Playing

# Convert video for rerun compatibility
python convert_videos_for_rerun.py Data/Videos/

Memory Issues with Large Datasets

# Reduce subsampling percentage
python visualize_slam_output.py ./outputs --subsample-percentage 0.3
# Or disable subsampling entirely
python visualize_slam_output.py ./outputs --no-subsampling

Missing Data Files

• Check that the output directory contains the expected SLAM analysis files
• Verify file naming conventions match the supported patterns
• Ensure metadata.json is present for automatic configuration

Performance Tips

• Use --subsample-percentage for large point clouds
• Disable unused visualizations with --no-* flags
• Enable --color_pointcloud_by_classIds for better segmentation visualization
• Use --show_segmentation_masks_separately for detailed class analysis

Configuration

Pipeline Metadata Integration

The script automatically reads pipeline configuration from metadata.json:

• View cone settings from IncrementalClosestPointFinderComponent
• Video paths from MAST3RSLAMVideoDataset
• File paths and analysis parameters

Default Settings

• Floor threshold: 0.05m
• Grid size: 2.0m
• View cone angle: 45°
• Subsampling: 50% for temporal steps
• Warning distances: 0.2m, 0.4m, 0.6m

Output

The visualizer creates an interactive 3D scene in the rerun web viewer with:

• World coordinate system: All spatial data in 3D space
• Timeline controls: Playback controls for temporal data
• Scene tree: Organized visualization hierarchy
• Real-time plots: Distance and warning level charts
• Text overlays: Warning messages and status information