ORB-SLAM3 uses YAML configuration files to define camera parameters, ORB extractor settings, IMU parameters, and visualization options. This guide explains the structure and parameters of these configuration files.
Configuration File Structure A typical ORB-SLAM3 configuration file contains several sections:
System Configuration : Atlas loading/saving settingsCamera Parameters : Intrinsics, distortion, and stereo/depth parametersIMU Parameters : Noise characteristics and transformations (for inertial sensors)ORB Extractor Settings : Feature detection parametersViewer Parameters : Visualization settings
Basic Configuration Template All configuration files start with the YAML version declaration:
System Configuration Optional settings for loading and saving map sessions:
#-------------------------------------------------------------------------------------------- # System config #-------------------------------------------------------------------------------------------- # Load a previous session (Atlas) #System.LoadAtlasFromFile: "Session_MH01_MH02_MH03_Stereo60_Pseudo" # Save the current session to a file #System.SaveAtlasToFile: "Session_MH01_MH02_MH03_Stereo60_Pseudo"
If the LoadFile doesn’t exist, the system will create a new Atlas from scratch. If a SaveFile already exists, it will be overwritten.
Camera Parameters Camera Type ORB-SLAM3 supports two camera models:
File.version : "1.0" # Choose camera model Camera.type : "PinHole" # For pinhole cameras # OR Camera.type : "KannalaBrandt8" # For fisheye cameras
Pinhole Camera Configuration # Camera calibration and distortion parameters (OpenCV) Camera1.fx : 517.306408 Camera1.fy : 516.469215 Camera1.cx : 318.643040 Camera1.cy : 255.313989 # Distortion coefficients Camera1.k1 : 0.262383 Camera1.k2 : -0.953104 Camera1.p1 : -0.005358 Camera1.p2 : 0.002628 Camera1.k3 : 1.163314 # Image dimensions Camera.width : 640 Camera.height : 480 # Frame rate Camera.fps : 30 # Color order (0: BGR, 1: RGB) Camera.RGB : 1
# Left camera parameters Camera1.fx : 458.654 Camera1.fy : 457.296 Camera1.cx : 367.215 Camera1.cy : 248.375 Camera1.k1 : -0.28340811 Camera1.k2 : 0.07395907 Camera1.p1 : 0.00019359 Camera1.p2 : 1.76187114e-05 # Right camera parameters Camera2.fx : 457.587 Camera2.fy : 456.134 Camera2.cx : 379.999 Camera2.cy : 255.238 Camera2.k1 : -0.28368365 Camera2.k2 : 0.07451284 Camera2.p1 : -0.00010473 Camera2.p2 : -3.55590700e-05 # Image dimensions Camera.width : 752 Camera.height : 480 # Frame rate Camera.fps : 20 # Color order Camera.RGB : 1 # Stereo baseline times depth threshold Stereo.ThDepth : 60.0 # Transformation from left to right camera Stereo.T_c1_c2 : !!opencv-matrix rows : 4 cols : 4 dt : f data : [ 0.999997256477797 , -0.002317135723275 , -0.000343393120620 , 0.110074137800478 , 0.002312067192432 , 0.999898048507103 , -0.014090668452683 , -0.000156612054392 , 0.000376008102320 , 0.014089835846691 , 0.999900662638081 , 0.000889382785432 , 0 , 0 , 0 , 1.000000000000000 ]
# Camera calibration parameters Camera1.fx : 517.306408 Camera1.fy : 516.469215 Camera1.cx : 318.643040 Camera1.cy : 255.313989 Camera1.k1 : 0.262383 Camera1.k2 : -0.953104 Camera1.p1 : -0.005358 Camera1.p2 : 0.002628 Camera1.k3 : 1.163314 # Image dimensions Camera.width : 640 Camera.height : 480 # Frame rate Camera.fps : 30 # Color order Camera.RGB : 1 # Depth parameters Stereo.ThDepth : 40.0 Stereo.b : 0.07732 # Baseline # Depth map scale factor RGBD.DepthMapFactor : 5000.0 # Use 1.0 for ROS bags
Parameter Descriptions
fx, fy: Focal lengths in pixels (x and y directions)cx, cy: Principal point coordinates (optical center) These parameters define the camera’s internal geometry and can be obtained through calibration.
Radial and tangential distortion parameters following OpenCV convention:
k1, k2, k3: Radial distortion coefficientsp1, p2: Tangential distortion coefficients For low-distortion cameras, some coefficients may be zero or omitted.
Stereo.ThDepth: Maximum depth threshold (baseline × threshold value)Stereo.b: Stereo baseline in meters (for RGB-D)Stereo.T_c1_c2: 4×4 transformation matrix from left to right camera The transformation matrix includes rotation and translation between stereo camera pairs.
RGBD.DepthMapFactor: Scale factor to convert depth values to meters
Use 5000.0 for TUM RGB-D dataset format (depth in mm × 5)
Use 1.0 if depth is already in meters (common for ROS)
IMU Parameters For visual-inertial configurations, add IMU parameters:
# Transformation from camera to body-frame (IMU) IMU.T_b_c1 : !!opencv-matrix rows : 4 cols : 4 dt : f data : [ 0.0148655429818 , -0.999880929698 , 0.00414029679422 , -0.0216401454975 , 0.999557249008 , 0.0149672133247 , 0.025715529948 , -0.064676986768 , -0.0257744366974 , 0.00375618835797 , 0.999660727178 , 0.00981073058949 , 0.0 , 0.0 , 0.0 , 1.0 ] # IMU noise characteristics IMU.NoiseGyro : 1.7e-04 # Gyroscope measurement noise (rad/s) IMU.NoiseAcc : 2.0e-03 # Accelerometer measurement noise (m/s^2) IMU.GyroWalk : 1.9393e-05 # Gyroscope random walk (rad/s^2) IMU.AccWalk : 3.0e-03 # Accelerometer random walk (m/s^3) IMU.Frequency : 200.0 # IMU sample rate (Hz)
IMU noise parameters should be obtained from your sensor’s datasheet or through Allan variance analysis. See the Calibration Guide for details. ORB feature extraction parameters control the number and quality of detected features:
#-------------------------------------------------------------------------------------------- # ORB Parameters #-------------------------------------------------------------------------------------------- # Number of features per image ORBextractor.nFeatures : 1200 # Scale factor between levels in the scale pyramid ORBextractor.scaleFactor : 1.2 # Number of levels in the scale pyramid ORBextractor.nLevels : 8 # FAST threshold # Image is divided in a grid. At each cell FAST features are extracted. # First iniThFAST is used. If no corners detected, minThFAST is used. ORBextractor.iniThFAST : 20 ORBextractor.minThFAST : 7
Parameter Guidelines Default Settings
Low-Texture Environments
High-Speed Motion
ORBextractor.nFeatures : 1200 ORBextractor.scaleFactor : 1.2 ORBextractor.nLevels : 8 ORBextractor.iniThFAST : 20 ORBextractor.minThFAST : 7
Target number of ORB features to extract per frame.
Lower values (500-800): Faster processing, less robust trackingHigher values (1500-2000): More robust, higher computational costRecommended : 1000-1200 for most applications
Scale factor between consecutive pyramid levels.
Typical values: 1.2 (more levels, finer scale) to 1.5 (fewer levels)
Affects scale invariance and computational cost
Recommended : 1.2 for general use
Number of pyramid levels for scale-space feature detection.
More levels = better scale invariance but higher computation
Recommended : 8 levels for most scenarios
Thresholds for FAST corner detection:
iniThFAST: Initial threshold tried firstminThFAST: Fallback threshold if insufficient features found Lower thresholds for low-contrast or poorly textured scenes.Viewer Parameters Visualization settings for the Pangolin viewer:
#-------------------------------------------------------------------------------------------- # Viewer Parameters #-------------------------------------------------------------------------------------------- Viewer.KeyFrameSize : 0.05 Viewer.KeyFrameLineWidth : 1.0 Viewer.GraphLineWidth : 0.9 Viewer.PointSize : 2.0 Viewer.CameraSize : 0.08 Viewer.CameraLineWidth : 3.0 Viewer.ViewpointX : 0.0 Viewer.ViewpointY : -0.7 Viewer.ViewpointZ : -1.8 Viewer.ViewpointF : 500.0 Viewer.imageViewScale : 1.0
These parameters control the 3D visualization appearance and can be adjusted for better visibility based on your scene scale.
Example Configurations EuRoC MAV Dataset (Stereo-Inertial) Complete configuration for EuRoC dataset with stereo cameras and IMU:
% YAML :1.0 File.version : "1.0" Camera.type : "PinHole" Camera1.fx : 458.654 Camera1.fy : 457.296 Camera1.cx : 367.215 Camera1.cy : 248.375 Camera1.k1 : -0.28340811 Camera1.k2 : 0.07395907 Camera1.p1 : 0.00019359 Camera1.p2 : 1.76187114e-05 Camera2.fx : 457.587 Camera2.fy : 456.134 Camera2.cx : 379.999 Camera2.cy : 255.238 Camera2.k1 : -0.28368365 Camera2.k2 : 0.07451284 Camera2.p1 : -0.00010473 Camera2.p2 : -3.55590700e-05 Camera.width : 752 Camera.height : 480 Camera.fps : 20 Camera.RGB : 1 Stereo.ThDepth : 60.0 Stereo.T_c1_c2 : !!opencv-matrix rows : 4 cols : 4 dt : f data : [ 0.999997256477797 , -0.002317135723275 , -0.000343393120620 , 0.110074137800478 , 0.002312067192432 , 0.999898048507103 , -0.014090668452683 , -0.000156612054392 , 0.000376008102320 , 0.014089835846691 , 0.999900662638081 , 0.000889382785432 , 0 , 0 , 0 , 1.000000000000000 ] IMU.T_b_c1 : !!opencv-matrix rows : 4 cols : 4 dt : f data : [ 0.0148655429818 , -0.999880929698 , 0.00414029679422 , -0.0216401454975 , 0.999557249008 , 0.0149672133247 , 0.025715529948 , -0.064676986768 , -0.0257744366974 , 0.00375618835797 , 0.999660727178 , 0.00981073058949 , 0.0 , 0.0 , 0.0 , 1.0 ] IMU.NoiseGyro : 1.7e-04 IMU.NoiseAcc : 2.0e-03 IMU.GyroWalk : 1.9393e-05 IMU.AccWalk : 3.0e-03 IMU.Frequency : 200.0 ORBextractor.nFeatures : 1200 ORBextractor.scaleFactor : 1.2 ORBextractor.nLevels : 8 ORBextractor.iniThFAST : 20 ORBextractor.minThFAST : 7
Creating Your Own Configuration
Calibrate your camera
Use the calibration tools and methods described in the Calibration Guide to obtain accurate intrinsic and distortion parameters.
Choose a template
Start with an example configuration file from Examples/ that matches your sensor type (Mono, Stereo, RGB-D, with or without IMU).
Update camera parameters
Replace the calibration values with your camera’s parameters. Ensure the image dimensions match your actual image size.
Tune ORB parameters
Start with default ORB settings and adjust based on your environment:
Lower FAST thresholds for low-texture scenes
Increase nFeatures for challenging environments
Test and validate
Run ORB-SLAM3 with your configuration and verify tracking performance. Check the viewer for feature distribution.
Incorrect camera calibration parameters will result in poor tracking and mapping performance. Always verify your calibration before use.
Next Steps 
