OpenDroneKit

Offline-first drone inspection toolkit for mission planning, data capture QA, defect analytics, reconstruction, and reporting.

What This Project Does

• Plans repeatable inspection missions (grid, facade, corridor, tower, solar, waypoints, linked missions).
• Applies mission safety constraints (geofence, no-fly zones, altitude bands, standoff, RTH rules).
• Supports terrain-aware planning (AGL/AMSL mode, terrain source input, slope-normal camera behavior, wind-aware speed adjustments).
• Runs post-flight analysis (coverage QA, crack/metal/structural/solar defect detection, crack propagation).
• Builds reconstruction artifacts (point cloud, mesh, orthomosaic, DSM/DTM, digital twin metadata).
• Generates project reports from run outputs.

UI Workflow

• Projects: create/select project, offline/sync status.
• Missions: map-based mission design + export.
• Fly: preflight and run controls.
• Data: dataset import and quick review.
• Analysis: operations, full pipeline, crack/propagation, reconstruction, defect models.
• Reports: standard/advanced report generation.

Quick Start

python main.py

CLI pipeline:

python run_pipeline.py --images <dataset_dir> --output final_toolkit_outputs

Repository Layout

main.py
run_pipeline.py
mission/
core/
ui/
models/
training/

Model Assets And Licensing

Model files and provenance:

• Registry: models/model_registry.json
• Provenance map: models/manifests/model_provenance.json

Current externally sourced model origins used in this repo:

1. training/sources/GBH-YOLOv5
   License file present: GPL-3.0 (training/sources/GBH-YOLOv5/LICENSE)
2. training/sources/crack-segmentation
   License file present: MIT (training/sources/crack-segmentation/LICENSE)
3. training/sources/Concrete-defect-detection
   No license file found in the cloned source tree at time of integration.

Important compliance note:

• If a model source has no explicit license, do not assume redistribution rights.
• Before distributing binaries, Docker images, or public model weights, verify and document rights for each model file.
• For uncertain model assets, replace them with models that have explicit, compatible licenses.

Outputs

By default, outputs are written under:

final_toolkit_outputs/

Typical run artifacts:

• summary.json
• inspection_report.md
• reconstruction outputs (.ply, mesh, orthomosaic, DSM/DTM, digital twin metadata)

Notes

• Cloud reconstruction mode currently logs request/response and uses local fallback in this build.
• GUI flight execution is simulated workflow control, not direct autopilot flight control.