Open Video Tracker — User Guide v0.2

• Download the Open Video Tracker add‑on (Windows only for now).
• If you have an NVIDIA RTX GPU, download the CUDA build; otherwise, download the standard build.
• Locate the open_video_tracker folder containing the add‑on files.

1. Open Blender
2. Go to Edit → Preferences
3. Select the Add-ons tab
4. Click Install from Disk
5. Navigate to and select the open_video_tracker folder
6. Click Install From Disk

1. Switch to the 3D Viewport
2. Open the right sidebar and look for the Open Video Tracker tab
3. If it’s visible, the installation was successful

The main interface is located in the 3D Viewport sidebar under the “Open Video Tracker” tab.

• Video Path: File browser to select input video file
• Video Information: Displays frame rate, resolution, and bitrate

• Quality: Native, High, Balanced, Low, Lowest

• Max Image Size (default: 2000 px)
• Camera Model: Simple Radial recommended
• Max Num Features (default: 8192)

v0.2 Update: GPU settings have been moved to Preferences for more fine-grained control. See Configuration section for details.

• Overlap: Number of overlapping frames (default: 10)

• Max Tracks per image (default: 1000)
• Constraint Type: Balance between point and camera constraints
• Advanced Options: GLOMAP parameters (epipolar error, iterations)

• Track Video: Start the processing pipeline
• Progress: Real-time progress indicator

v0.2 Enhancement: The import menu has been redesigned with improved organization and visual feedback.

Camera Import

• Import Cameras: Enable camera track import
• Camera Extent: Size of camera visualization
• Add Background Images
• Add Image Planes
• Animation Options

Point Import

• Import Points
• Point Cloud Display: Sparsity and rendering options
• GPU Rendering
• Mesh Options

Advanced Options (v0.2)

• 🔄 Reimport Data: New in v0.2 - Import previously processed data without reprocessing
• 📐 Z-Axis Postprocessing: New in v0.2 - Automatic correction for Z-axis alignment issues

• Supported formats: MP4, AVI, MOV, MKV, WMV, FLV, WebM
• Use high-quality, stable footage
• Ensure good lighting and contrast
• Start with short clips for testing

Video Quality

• Native/High: Best quality, slowest processing
• Balanced: Good quality/performance
• Low/Lowest: Fastest, smaller files

Feature Extraction

• Increase Max Image Size for high-res videos
• Enable GPU if available
• Choose camera model based on camera type

Reconstruction

• Adjust Max Tracks for scene complexity
• Use Balanced constraints for most cases

• Click Track Video
• Monitor progress
• Processing time depends on video length and settings
• Do not close Blender during processing

After processing completes, the add‑on automatically imports:

• Camera objects with animation data
• Point cloud data
• Background images (if enabled)

• Enable GPU Acceleration: Use CUDA for NVIDIA RTX GPUs
• GPU Memory Limit: Set maximum GPU memory usage
• Compute Device: Select specific GPU device (multi-GPU systems)
• Force CPU Fallback: Override GPU settings for troubleshooting

GPU settings are now accessible via Edit → Preferences → Add-ons → Open Video Tracker → GPU Configuration

• Simple Radial: Most consumer cameras (smartphones, DSLRs)
• Simple Radial Fisheye: Action cameras (GoPro)
• OpenCV: Computer vision cameras
• Full OpenCV: Complex distortion scenarios

• Points Only: When camera positions are unreliable
• Cameras Only: When point features are sparse
• Balanced: Recommended default
• Points and Cameras: Maximum constraints, slower

• Z-Axis Auto-Correction: Automatically fix Z-axis alignment issues
• Scale Normalization: Normalize point cloud scale
• Outlier Removal: Remove statistical outliers from point clouds
• Smoothing: Apply mesh smoothing for better surface quality

• Each step is logged to the Blender console
• View via Window → Toggle System Console (Windows)

The add‑on uses the same directory as the .blend file and creates:

{your_blend_file_directory}/
└── video_tracking/
    └── {video_name}/
        ├── images/           # Extracted frames
        ├── sparse/           # Reconstruction data
        │   ├── 0/            # Model files
        │   ├── cameras.txt   # Camera data
        │   ├── images.txt    # Image data
        │   └── points3D.txt  # Point cloud data
        └── database.db       # COLMAP database

• Quick Reimport: Import previously processed results without reprocessing
• Selective Reimport: Choose specific data types to reimport
• Update Existing: Update current scene without creating duplicates
• History Browser: Browse and select from previous processing sessions

• Camera Extent: Visual size of camera objects
• Background Images: Show original frames
• Image Planes: Create textured planes
• Animation: Convert camera motion to keyframes

• Display Sparsity: Reduce density for performance
• GPU Rendering
• Mesh Conversion: Convert points to mesh objects

• Auto Z-Correction: Automatically detect and fix Z-axis issues
• Manual Z-Offset: Apply custom Z-axis offset if needed
• Ground Plane Detection: Automatically detect and align to ground plane
• Preview Mode: Preview corrections before applying

• Automatic Detection: Identifies common Z-axis alignment issues
• Smart Correction: Applies appropriate fixes based on scene analysis
• Preview and Adjust: See changes before applying
• Multiple Correction Methods: Ground plane alignment, scale-based correction, manual offset

• Outlier Removal: Statistical filtering of erroneous points
• Density Optimization: Balance point density for performance
• Noise Reduction: Smooth irregular surfaces
• Scale Normalization: Ensure consistent scale across imports

• Resolution: Higher resolution improves tracking but increases time
• Frame Rate: 24–30 fps is optimal
• Stabilization: Use stabilized footage when possible
• Lighting: Ensure consistent, even lighting
• Motion: Prefer smooth camera movement

• Start Small: Test with short clips first
• GPU Usage: Enable when available
• Quality Settings: Use Balanced for initial tests
• Memory Management: Monitor RAM with large videos

• Feature‑Rich scenes track better
• Minimize motion blur
• Include objects of known scale
• Ensure sufficient frame overlap