NodeODM and OpenDroneMap turn raw aerial photographs into orthomosaics, point clouds, elevation models, and 3D meshes — running entirely on your hardware. No cloud uploads. No subscription fees. No image limits. This guide walks you through the complete setup and workflow from flight to deliverable.
1 — Start NodeODM
docker run -d -p 3000:3000 --name nodeodm --restart unless-stopped opendronemap/nodeodm
2 — Verify it's running
curl http://localhost:3000/info
3 — Connect in mapplot Processing
http://localhost:3000
OpenDroneMap (ODM) is a free, open-source photogrammetry engine that reconstructs 3D geometry and georeferenced maps from sets of overlapping aerial photographs. Originally developed for academic research, it has grown into one of the most capable and widely adopted drone image processing systems used in professional surveying, mapping, and GIS workflows worldwide.
NodeODM is the REST API server that wraps the ODM engine. It exposes a clean HTTP interface for submitting image jobs, monitoring progress, and downloading outputs — making it straightforward to integrate drone processing into any application. mapplot connects directly to NodeODM running on your machine, handling job management, progress tracking, and output downloads within your existing project workflow.
Unlike commercial cloud processing platforms — Pix4D Cloud, DroneDeploy, Metashape Cloud — NodeODM runs entirely on your hardware. Images never leave your network. There are no per-image charges, no monthly processing credits, and no dependency on internet connectivity or third-party uptime.
The full stack — ODM engine and NodeODM API — is open source under the AGPL licence, maintained by the OpenDroneMap organisation and a global community of contributors. Commercial support is available from Dronesmith and other providers.
Complete data privacy
Imagery stays on your hardware from capture to delivery. Non-negotiable for government, defence, infrastructure, and commercially sensitive site work.
No upload bottleneck
Large datasets — 500+ images, RAW files, multi-spectral — take hours to upload to cloud platforms. Local processing starts in seconds.
Fully offline capable
Process in the field, on a site network, or anywhere with no internet connection. Only the initial Docker pull requires internet.
Zero processing fees
NodeODM is free. Process 50 images or 5,000 on the same job — the cost is your electricity and hardware, nothing more.
Full algorithm control
Access every OpenDroneMap option — feature matching algorithms, reconstruction quality, point cloud density, output resolution, and GCP integration.
Runs on your network
Deploy NodeODM on a local server and let multiple team members submit jobs without each needing their own processing machine.
How local processing with NodeODM stacks up against the major commercial platforms — cost, privacy, capability, and control.
| Feature | NodeODM + mapplot | Pix4D Cloud | DroneDeploy | WebODM (hosted) |
|---|---|---|---|---|
| Cost | Free (open source) | $350/mo+ | $299/mo+ | Free (self-hosted) |
| Images leave your machine | Never | Always | Always | Self-hosted: no |
| Works fully offline | Yes | No | No | Self-hosted: yes |
| Per-image / credit fees | None | Yes | Yes | None |
| Orthomosaic (GeoTIFF) | Yes | Yes | Yes | Yes |
| Dense point cloud (LAS/LAZ) | Yes | Yes | Yes | Yes |
| 3D textured mesh (OBJ/GLB) | Yes | Yes | No | Yes |
| Multispectral / NDVI outputs | Yes | Yes | Yes | Yes |
| GPU acceleration | Yes (CUDA) | Cloud-side | Cloud-side | Yes (self-hosted) |
| DJI KMZ mission integration | Via mapplot | No | Yes | No |
| Open source | Yes (AGPL) | No | No | Yes (AGPL) |
Pricing as of 2025. NodeODM is free and open source — always. Cloud platform costs vary by plan and volume. WebODM self-hosted is also free and uses the same ODM engine as NodeODM.
A single NodeODM job can generate all six output types simultaneously from the same image set. Select only the ones you need — each is a standard format compatible with professional GIS, CAD, and BIM tools.
GeoTIFF (.tif)A geometrically corrected, georeferenced aerial photograph stitched from hundreds of overlapping drone images. Every pixel has a real-world coordinate. The primary deliverable for land survey, construction monitoring, agriculture, and most mapping workflows.
GeoTIFF (.tif)A raster elevation model representing the height of the highest surface at each point — buildings, tree canopy, and all above-ground features included. Essential for volume calculations, line-of-sight analysis, and solar energy assessments.
GeoTIFF (.tif)Bare-earth elevation with vegetation and structures algorithmically removed. Used where the true ground surface matters — slope analysis, corridor surveying, hydrological modelling, and civil engineering design.
LAS / LAZ / PLYMillions of 3D coordinate points derived from photogrammetric reconstruction, each with XYZ position and RGB colour from the imagery. The foundation for BIM integration, heritage recording, infrastructure inspection, and forestry analysis.
OBJ / GLB / FBXA photorealistic 3D model with full colour texture mapped from your drone imagery. Ideal for client presentations, augmented reality deliverables, asset documentation, and embedding in web viewers like Cesium or Sketchfab.
GeoJSON / DXF / SHPElevation isolines derived from the DSM or DTM at any specified vertical interval — 0.25 m, 0.5 m, 1 m, or custom. Ready to import into AutoCAD, ArcGIS, QGIS, or MicroStation for topographic mapping and civil design.
Seven stages from mission planning to client-ready outputs — each one handled by mapplot and NodeODM working together.
Draw your survey boundary using mapplot's polygon, rectangle, or corridor tools on the satellite map. Configure altitude (AGL), flight speed, gimbal pitch, front overlap, and side overlap. mapplot generates optimised flight lines and previews photo footprints so you can verify area coverage before committing to a flight. Export the KMZ file directly to your drone controller.
Tips
Execute the planned mission from your drone controller. Fly in consistent, diffuse lighting — overcast days produce the most uniform results. Avoid strong shadows, specular reflections on water, and high winds that cause camera shake. The camera must embed GPS coordinates and altitude in EXIF data — verify this before your first job.
Tips
Copy all images from the SD card into a single dedicated folder on your computer. Keep each flight as a separate folder — one folder equals one processing job. Remove any images taken during takeoff, landing, or outside the survey area. More images in the job means longer processing, so only include the ones that matter.
Tips
NodeODM runs inside a Docker container — a self-contained package that includes the entire OpenDroneMap engine and all its dependencies. Install Docker Desktop first (a one-time step), then start NodeODM with a single command. The first pull downloads approximately 1 GB. After that, starting NodeODM takes under 10 seconds.
Tips
In mapplot, navigate to Processing. If NodeODM is not yet connected, you'll be guided through the connection step. Start a new job, name it, upload your images from the folder you prepared, select a quality preset, and enable the output types you need. Click Start — NodeODM handles the rest automatically: feature extraction, sparse reconstruction, dense cloud, and all outputs.
Tips
Processing runs entirely on your machine with no cloud involvement. mapplot polls NodeODM every 15 seconds for live progress updates and displays a progress bar per job. A 300-image survey at medium quality on a modern 8-core CPU typically finishes in 30–60 minutes. GPU acceleration can reduce this by 3–5×. Multiple jobs queue automatically.
Tips
Once processing completes, download your selected outputs from the mapplot job detail panel. Load orthomosaics into QGIS or ArcGIS for measurement and analysis, import point clouds into CloudCompare or Autodesk ReCap, and share 3D meshes via Sketchfab or the mapplot client portal. Every byte of data stayed on your hardware throughout the entire process.
Tips
NodeODM runs on any modern computer capable of running Docker Desktop. RAM is the most important factor — more is always better for large datasets.
Three steps from a clean machine to a running NodeODM instance. The entire setup takes under 15 minutes including the Docker image download.
Docker Desktop is available for Windows, macOS, and Linux. Download the installer for your OS, run it, and start Docker Desktop. Wait for the engine to fully initialise — the whale icon in your system tray turns steady (not animated) when it's ready.
Download Docker DesktopOption A — Standalone (recommended for most users)
docker run -d -p 3000:3000 --name nodeodm --restart unless-stopped opendronemap/nodeodm
The first run pulls the NodeODM image (~1 GB). Subsequent starts take under 10 seconds. The --restart unless-stopped flag makes NodeODM start automatically whenever Docker starts.
Option B — With mapplot docker-compose (if you have the repo)
docker compose up -d
Starts all mapplot services including NodeODM in one command from the project root directory.
Verify NodeODM is running
curl http://localhost:3000/info
You should see a JSON response like {"version":"x.x.x","taskQueueCount":0}. You can also open http://localhost:3000 directly in a browser.
In mapplot, navigate to Processing. If NodeODM is not yet connected, the setup screen will guide you through it. Enter http://localhost:3000 (or your server's address), click Test Connection, and mapplot will verify and save your URL automatically. You're ready to process.
An NVIDIA GPU can reduce processing time by 3–5×. GPU support requires the NVIDIA Container Toolkit on Linux, or WSL2 with CUDA on Windows. AMD GPUs are not currently supported in Docker.
Install NVIDIA Container Toolkit (Ubuntu/Debian):
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list && sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit && sudo nvidia-ctk runtime configure --runtime=docker && sudo systemctl restart docker
Start NodeODM with GPU passthrough:
docker run -d -p 3000:3000 --name nodeodm --gpus all --restart unless-stopped opendronemap/nodeodm
Windows users: see the NVIDIA WSL2 CUDA guide for enabling GPU passthrough through WSL2. Requires Windows 11 or Windows 10 21H2+.
NodeODM offers five quality presets that trade processing time against output resolution and reconstruction accuracy. Medium is the right starting point for the majority of drone survey work.
Verify alignment before committing to a full run. Catch coverage gaps fast.
Quick site overview and proof-of-concept deliverables where fine detail isn't needed.
Standard mapping, construction monitoring, and most client deliverables. Start here.
Survey-grade orthomosaics and engineering datasets where GSD and accuracy are critical.
Heritage recording, detailed inspection reports, and research requiring maximum fidelity.
Curated YouTube channels covering OpenDroneMap setup, photogrammetry processing, output analysis, and GCP workflows. All links open directly on YouTube.
OpenDroneMap
Release demos, community case studies, processing tutorials, and feature walkthroughs from the ODM core team.
Drone Launch Academy
Practical walkthroughs covering drone mapping software, photogrammetry workflows, and client deliverable creation.
Dylan Gorman
In-depth tutorials on GCP integration, accuracy analysis, and survey-grade processing from a working photogrammetrist.
QGIS Official
How to import, georeference, analyse, and export orthomosaics and elevation models in QGIS.
Emlid
RTK GNSS measurement, GCP placement, and integration with photogrammetry processing for centimetre-level accuracy.
Pilot Institute
Accessible guides to planning mapping missions, processing results, and delivering professional outputs.
Everything you need to go further — official docs, output viewers, the OpenDroneMap community, and specialist tools for working with your processed data.
OpenDroneMap Documentation
Complete reference for all ODM options, output formats, GCP workflow, and troubleshooting
NodeODM GitHub Repository
Source code, REST API reference, changelog, and issue tracker
ODM Core Engine (GitHub)
The open-source photogrammetry engine that powers NodeODM — full algorithm documentation
ODM Options Reference
Every processing option explained — quality, feature algorithms, point cloud density, and more
Docker Desktop Download
Free Docker Desktop for Windows, macOS, and Linux — required to run NodeODM
NodeODM on Docker Hub
Official image — tagged releases and latest build information
NVIDIA Container Toolkit
Required for GPU acceleration — enables CUDA passthrough into Docker containers
WSL2 CUDA Setup (Windows)
NVIDIA guide for enabling GPU support inside WSL2 Docker on Windows
OpenDroneMap Community Forum
Active community — processing questions, workflow tips, troubleshooting, and hardware advice
WebODM (GUI for NodeODM)
Full-featured web interface built on NodeODM with project management and team features
ODM GitHub Discussions
Feature requests, algorithm deep-dives, and community Q&A with the development team
QGIS — Free GIS Platform
Industry-standard open-source GIS for viewing orthomosaics, DSMs, contours, and vector outputs
CloudCompare — Point Cloud
Free point cloud viewer and analysis tool — opens LAS, LAZ, PLY, and E57 files
ExifTool — EXIF Inspector
Verify GPS coordinates are embedded in your images before submitting a processing job
Potree — Web Point Cloud Viewer
Open-source WebGL point cloud renderer for delivering large point clouds to clients via browser
LAStools — LAS/LAZ Processing
Professional command-line tools for classifying, filtering, and converting point cloud files
Common questions about NodeODM setup, processing performance, hardware requirements, image quality, and workflow integration.
Processing time depends on image count, quality preset, and hardware. A 200-image dataset at Medium quality typically finishes in 30–45 minutes on a modern 6-core CPU. High quality with 500 images can take several hours. An NVIDIA GPU reduces time by 3–5×. NodeODM scales near-linearly with CPU cores — a 16-core machine processes the same job in roughly half the time of an 8-core machine.
There is no hard limit. Consumer hardware handles 200–1,000 images comfortably on Medium quality. Datasets above 2,000 images benefit from 32 GB+ RAM and a capable GPU. For very large survey areas, split missions into sub-areas using mapplot's drawing tools — process each as a separate job, then mosaic the results in QGIS using the Merge Rasters tool.
A GPU is not required — NodeODM processes entirely on CPU with no degradation in output quality. However, an NVIDIA GPU with CUDA support significantly accelerates feature extraction and dense point cloud generation (the two most time-consuming stages). AMD GPUs have limited Docker support. If you process drone images more than a few times a week, an NVIDIA RTX 3060 or better is a worthwhile investment.
OpenDroneMap recommends a minimum of 70% front (along-track) overlap and 60% side (across-track) overlap for 2D mapping. For reliable 3D reconstruction, use 80%/70% or more. mapplot calculates and previews photo footprints at any overlap setting before you fly. Insufficient overlap is the most common cause of failed or partial reconstructions — when uncertain, fly with more overlap.
NodeODM accepts JPEG and TIFF. Most consumer drone cameras output JPEG, which works well for photogrammetry. For maximum radiometric depth — particularly useful for NDVI analysis or precise colour correction — shoot RAW and convert to 16-bit TIFF before processing. Every image must contain GPS coordinates in its EXIF metadata; without GPS, NodeODM cannot georeference any outputs.
Yes. NodeODM is a REST API and runs anywhere Docker runs — on a local Linux server, high-spec NAS (Synology, QNAP), cloud VM (AWS, GCP, Azure), or a Raspberry Pi (very slow). In mapplot, you can point the NodeODM URL to any IP address on your local network, allowing one powerful machine to serve multiple users. For remote access, expose port 3000 securely with a reverse proxy.
NodeODM is the headless REST API server — it receives image uploads, runs the ODM engine, and serves outputs. It has no graphical interface. WebODM is an open-source web application built on top of NodeODM that adds a GUI, project management, and team collaboration features. mapplot connects directly to NodeODM, so you get full processing power integrated into your mapplot workflow without needing a separate WebODM installation.
Stop the running container, pull the latest image, and restart: docker stop nodeodm && docker rm nodeodm && docker pull opendronemap/nodeodm && then run your start command again. If you use docker compose, run docker compose pull followed by docker compose up -d. Updates typically bring improved algorithms, bug fixes, and new output options.
Yes. NodeODM processes multispectral imagery from cameras including MicaSense RedEdge-MX, Parrot Sequoia, DJI Multispectral, and Sentera. It produces NDVI, NDRE, VARI, and other vegetation index outputs. Accurate radiometric results require calibration panel images captured before and after each flight. The processing workflow is identical — upload your images and select the relevant output options.
Check the job log in mapplot for error messages — most failures have a clear cause. Common issues: insufficient image overlap (reconstruction fails to close), missing GPS EXIF tags (no georeferencing), insufficient RAM (process killed mid-job), or Docker not running when the job is submitted. For poor but completed results, run at a higher quality preset or add Ground Control Points (GCPs). The OpenDroneMap community forum is an excellent resource for diagnosing unusual errors.
Ground Control Points (GCPs) are physical targets placed on the ground and measured with a GNSS receiver (GPS or RTK). Including GCPs in your processing job improves absolute accuracy from 1–3 m (GPS-only drone) to centimetre level. They are required for survey-grade deliverables and engineering applications. mapplot's GCP module helps you place and export GCP coordinates. NodeODM accepts GCP files in the standard format during processing.
Yes — NodeODM exposes a clean REST API at localhost:3000. You can POST images, poll for status, and download outputs programmatically. The NodeODM API documentation on GitHub documents every endpoint. mapplot's processing pipeline uses this API internally. For automation, you can also use the official nodeodm npm package or the Python client library.
Can't find what you're looking for?
The OpenDroneMap community forum is active and welcoming — most questions are answered within hours. For mapplot-specific issues, contact our support team.
The most frequently encountered problems when setting up and running NodeODM — and exactly how to solve them.
Processing fails immediately with "out of memory" error
Cause
Docker RAM allocation too low — the default 2 GB is insufficient for any meaningful dataset.
Fix
Open Docker Desktop → Settings → Resources → Memory. Increase to at least 8 GB (16 GB recommended). Apply and restart Docker.
Job fails with "could not find enough matches" or SfM reconstruction fails
Cause
Insufficient image overlap — too few matching features between adjacent frames.
Fix
Re-fly with 75% front and 70% side overlap minimum. Ensure the area has surface texture — glass facades, water, and uniform sand all cause matching failures regardless of overlap.
Outputs not georeferenced — coordinates appear in the ocean or at 0,0
Cause
Images are missing GPS EXIF data — the most common silent processing failure.
Fix
Enable GPS logging in your drone camera settings before flying. Verify with ExifTool: exiftool -GPSLatitude image.jpg. Images without GPS cannot be georeferenced without GCPs.
NodeODM container fails to start — "port is already allocated" error
Cause
Another service is already bound to port 3000 on your machine.
Fix
Use a different host port: docker run -d -p 3001:3000 --name nodeodm opendronemap/nodeodm. Then update mapplot to connect to http://localhost:3001.
Orthomosaic has visible seams, colour banding, or blurry patches
Cause
Inconsistent lighting between images, or insufficient overlap in the affected area.
Fix
Fly in consistent overcast light. Avoid midday sun with hard shadows. Reduce flight speed to minimise motion blur. Check mapplot's photo preview for coverage gaps.
Processing is much slower than expected — taking many hours on a small dataset
Cause
CPU-only processing with restrictive Docker resource limits, or wrong quality preset selected.
Fix
Check Docker Desktop → Resources — ensure NodeODM can use all CPU cores. Switch from High/Ultra to Medium for faster validation. An NVIDIA GPU reduces time by 3–5×.
mapplot shows "connection failed" — cannot reach NodeODM
Cause
NodeODM container has stopped, wrong port, or Windows Firewall blocking localhost.
Fix
Run docker ps to check if nodeodm is in "Up" state. If stopped: docker start nodeodm. Verify the port in mapplot matches your NodeODM start command. On Windows, check Defender Firewall is not blocking port 3000.
mapplot handles the job management, progress tracking, and output downloads. You provide the drone images — NodeODM does the rest, on your own hardware.
No credit card required · Free plan includes 3 projects · Cancel anytime