How hussainmir05/3d-camera-depth-accuracy-evaluation works
This repository evaluates 3D camera depth accuracy from RGB images and point clouds. The documented workflow manually defines stair ROIs on RGB images, crops corresponding point-cloud regions, fits planes, and compares plane spacing with ground truth. A parallel analysis computes point-to-plane residual metrics to characterize surface noise. The README documents the stages and reported metrics; source excerpts were unavailable, so the graph reflects documented relationships only.
Main entry points: camera_test.py SVD Plane Fits
Components 12
ROI Segmentation
- Manual Stair ROIs
camera_test.py - ROI Point Crops
camera_test.py
Plane Geometry
- SVD Plane Fits
camera_test.py - Plane Spacing
camera_test.py - Point-Plane Residuals
camera_test.py
Accuracy Evaluation
- Spacing Error
camera_test.py - Surface Noise Metrics
camera_test.py - Evaluation Results
camera_test.py
Other components
- Operator
- RGB Images
- Point Clouds
- Ground Truth
Connections 12
- Operator → Manual Stair ROIs: defines regions
- RGB Images → Manual Stair ROIs: guides segmentation
- Point Clouds → ROI Point Crops: provides points
- Manual Stair ROIs → ROI Point Crops: selects regions
- ROI Point Crops → SVD Plane Fits: fits planes
- SVD Plane Fits → Plane Spacing: measures spacing
- SVD Plane Fits → Point-Plane Residuals: computes residuals
- Plane Spacing → Spacing Error: compares distances
- Ground Truth → Spacing Error: provides reference
- Point-Plane Residuals → Surface Noise Metrics: summarizes error
- Spacing Error → Evaluation Results: reports RMSE
- Surface Noise Metrics → Evaluation Results: reports MAE and STD