Skip to content

Options Selection Guide

This guide helps you choose the most appropriate WebODM processing parameters based on your drone, survey characteristics, and expected outputs. Please note: this guide is not perfect; finding the right parameters is not always a deterministic process—more often than not, it is an art that transcends algorithms.


Does your drone use a Global Shutter or a Rolling Shutter?

Section titled “Does your drone use a Global Shutter or a Rolling Shutter?”

If you are using a consumer drone (DJI Mini, DJI Air, Mavic, etc.)

Section titled “If you are using a consumer drone (DJI Mini, DJI Air, Mavic, etc.)”

Recommended setting

  • rolling-shutter: true Why? Rolling shutter sensors introduce geometric distortion when images are captured while the drone is moving (‘doming’ effect: the ground appears unnaturally arched). Best practice For future surveys, if possible:
  • reduce the flight speed;
  • use Stop-and-Hover mode when taking photographs.

Rolling-shutter is on but you notice ‘doming’ effects (where the ground appears unnaturally arched) and want to maximize the precision of the camera positioning?

Section titled “Rolling-shutter is on but you notice ‘doming’ effects (where the ground appears unnaturally arched) and want to maximize the precision of the camera positioning?”

Enable use-hybrid-bundle-adjustment. This increases the frequency of global and local optimization cycles during the SFM phase

Do you have a pre-existing, precise calibration file for your lens?

Section titled “Do you have a pre-existing, precise calibration file for your lens?”

Enable use-fixed-camera-params. This prevents the software from attempting to auto-calibrate the optical parameters

Are you using a multispectral or thermal sensor (e.g., for precision agriculture or temperature analysis)?

Section titled “Are you using a multispectral or thermal sensor (e.g., for precision agriculture or temperature analysis)?”

Note: read [Multispectral and Thermal section] (multispectral.md)

Once the process is complete, use the “Plant Health” tab in the map view to apply algorithms like NDVI, or the “Thermal” tab to view heat maps.


Do you want the software to automatically restrict the reconstruction to the specific perimeter where the drone photos were taken, avoiding unnecessary background processing?

Section titled “Do you want the software to automatically restrict the reconstruction to the specific perimeter where the drone photos were taken, avoiding unnecessary background processing?”

Enable auto-boundary. This creates a polygon around camera positions to limit the reconstruction area. You can refine the width of this boundary using auto-boundary-distance

auto-boundary: true

Recommended setting

sky-removal: true
  • Less reconstruction noise
  • Cleaner point cloud
  • Better mesh quality

Use:

bg-removal:true

Are you processing image frames extracted from a video file?

Section titled “Are you processing image frames extracted from a video file?”
  • video-limit: Defines the maximum number of frames to extract from the video. The default value is 500. Increasing this value can improve coverage, but it also proportionally increases processing time and memory (RAM) usage.
  • video-resolution: Sets the resolution (in pixels) of the extracted frames. For example, if a 4K video (3840×2160) is processed with this parameter set to 2000, the extracted frames will have a resolution of 2000×1125 pixels.

matcher-order: Perform image matching with the nearest N images based on image filename order. Can speed up processing of sequential images, such as those extracted from video. It is applied only on non-georeferenced datasets.

This is a key parameter when processing videos. Since the extracted frames are sequential, forcing the software to match images based only on their temporal order (for example, by setting a low value of 10 or 20) prevents every frame from being compared with all the others, saving hours of unnecessary computation.

‘.srt’ File: If you are using a DJI drone, the video is often accompanied by a subtitle file with the same name (for example, video.mp4 and video.srt). If you upload both files, WebODM will use the .srt file to extract GPS data and correctly associate it with the extracted frames.

  • Resize Images: The standard Resize Images option available in the task settings does not affect video files. To reduce the size of the frames extracted from a video, you must use the video-resolution parameter described above.

  • Quality: WebODM automatically filters out frames that are too dark or blurry during extraction to improve the quality of the final reconstruction.

  • Restart: If you change video-limit or video-resolution and want to restart the processing workflow, you will need to start again from the Load Dataset stage.

  • If you need to display the video on the map as supporting documentation (without processing it for photogrammetry), you can upload it using the Media button associated with an existing completed task


Does the area contain vegetation or low-texture surfaces?

Section titled “Does the area contain vegetation or low-texture surfaces?”

Examples:

  • forests
  • grasslands
  • agricultural fields
  • sand
  • snow

Recommended setting

min-num-features: 20000
  • increases the number of detectable feature points;
  • improves reconstruction robustness;
  • increases processing time.

Does the project include buildings or vertical structures?

Section titled “Does the project include buildings or vertical structures?”

Recommended settings

pc-quality: high
  • More accurate orthophotos
  • Better reconstruction of vertical walls
  • Sharper building edges

Do you need to preserve the camera’s full native resolution?

Section titled “Do you need to preserve the camera’s full native resolution?”

When flying very low (GSD below approximately 2 cm), WebODM may automatically reduce the processing resolution. To disable this optimization:

ignore-gsd: true

This significantly increases:

  • RAM usage
  • Disk space
  • Processing time

Main parameter:

mesh-octree-depth

Recommended values:

ScenarioValue
Flat terrain6–8
General purpose11
Complex architecture12

If increasing this value, also increase:

mesh-size

to avoid excessive mesh simplification.

Do you need a 3D mesh that is geometrically complete even in areas ‘unseen’ by the cameras (closed faces without color)?

Section titled “Do you need a 3D mesh that is geometrically complete even in areas ‘unseen’ by the cameras (closed faces without color)?”

Enable texturing-keep-unseen-faces.


Enable:

dtm: true

Adjust these parameters:

ParameterRecommendation
smrf-slope0.1 for flat terrain, up to 1.2 for mountainous terrain
smrf-thresholdMinimum object height to remove

Do you have an extremely dense point cloud and want to speed up DSM/DTM generation by simplifying the source data?

Section titled “Do you have an extremely dense point cloud and want to speed up DSM/DTM generation by simplifying the source data?”

Use dem-decimation. For example, a value of 50 tells the software to use only 2% of the points, significantly reducing file writing times

Does your orthophoto or DEM have jagged edges or ‘dragging’ artifacts at the borders that you want to trim away cleanly?

Section titled “Does your orthophoto or DEM have jagged edges or ‘dragging’ artifacts at the borders that you want to trim away cleanly?”

Set the crop parameter (in meters) to shrink the final boundary and remove interpolation artifacts.

Does your terrain model have ‘holes’ (areas without data) that you want to fill more accurately using multiple interpolation passes?

Section titled “Does your terrain model have ‘holes’ (areas without data) that you want to fill more accurately using multiple interpolation passes?”

Increase the dem-gapfill-steps to control the number of iterations the algorithm uses to fill empty cells.


Enable:

fast-orthophoto: true

Skips dense MVS reconstruction and generates the orthophoto directly from the sparse point cloud.


When the main goal is to obtain a result as quickly as possible (for example during emergency response, rapid assessment, field verification, or preliminary analysis), apply the following optimizations.

The fastest processing starts with an optimized acquisition strategy:

  • Perform a planar nadir flight whenever possible.
  • Maintain a constant altitude above ground.
  • Use regular image spacing and consistent overlap.
  • Avoid unnecessary oblique images if a 2D orthophoto is the primary goal.
  • Avoid capturing large areas outside the survey boundary.
  • Use a slower and more stable flight path when using rolling shutter cameras.

fast-orthophoto: true

Generates the orthophoto without performing the complete dense reconstruction phase.


resize-to: 2048

or a lower value depending on the required output quality.

Benefits:

  • significantly reduced processing time;
  • lower RAM consumption;
  • faster feature matching.

Generate only the products required for the task.

Avoid producing:

  • dense point cloud (the most important!)
fast-orthophoto: true
  • textured 3D mesh;
skip-3dmodel: true
  • (if unnecessary) DEM products.
dsm: false
dtm: false
  • report
skip-report:true

Example:

  • Emergency mapping → Orthophoto only
  • Preliminary inspection → Orthophoto + low-resolution DSM
  • Final survey → Full processing workflow

(only if you need the point cloud) Use:

pc-quality: low

or

pc-quality: medium

when a detailed 3D model is not required.

Benefits:

  • faster reconstruction;
  • lower disk usage.

If a mesh is not required:

mesh: false

Avoiding mesh generation can save a significant amount of processing time.


For maximum speed:

  1. Upload only the required images.
  2. Remove blurred or duplicated images before processing.
  3. Avoid images with large amounts of sky or irrelevant background.
  4. Use an area-based workflow instead of processing very large datasets at once.
  5. Split very large surveys into smaller independent blocks when possible.
  6. Use local processing hardware with GPU acceleration when available.

Enable:

build-overviews: true

Recent versions of ODX already generate Cloud Optimized GeoTIFFs (COGs) if you use the --cog option, which already include internal overviews.


Enable:

texturing-single-material: true
gltf: true
  • Easier import
  • Single texture material
  • Modern compressed 3D format

Enable:

3d-tiles: true

This generates optimized 3D Tiles suitable for web streaming.


Step 7 - Accuracy Verification / Alignment and Multitemporal Surveys

Section titled “Step 7 - Accuracy Verification / Alignment and Multitemporal Surveys”

Are Ground Control Points (GCPs) available?

Section titled “Are Ground Control Points (GCPs) available?”

To obtain an independent accuracy assessment:

  1. Select some control points as checkpoints.
  2. Prefix their names with:
CHK-

Checkpoint observations:

  • do not influence the bundle adjustment;
  • are used exclusively to compute independent accuracy statistics in the Quality Report.

Are you conducting a follow-up survey of an area processed previously, and do you need the new model (point cloud or DTM) to align perfectly with the old one?

Section titled “Are you conducting a follow-up survey of an area processed previously, and do you need the new model (point cloud or DTM) to align perfectly with the old one?”
Action: Use the align parameter. In the WebODM interface, select the reference task from the "Alignment" field. This instructs the software to ignore the original GPS/GCP data of the new survey and anchor it geometrically to the existing reconstruction.

SituationRecommended Parameter
Consumer dronerolling-shutter: true
RTK dronegps-accuracy
RTK + GCPforce-gps: true
Images include skysky-removal: true
Dense vegetationmin-num-features: 20000
Better building edgespc-quality: high
Very low GSDignore-gsd: true
Generate DTMdtm: true
Fast orthophotofast-orthophoto: true
Blender exportgltf: true
Cesium export3d-tiles: true
Accuracy validationCHK- checkpoints