Continuing our series of comparing Pix4D and WebODM, in this post we explore 2D outputs from a dataset covering an area containing multiple buildings and other structures.
Dataset
As before, we take a dataset to be what Pix4D has decided to showcase itself: a demo project from Pix4D. We will be using the UAV Demo Dataset as present on https://cloud.pix4d.com/demo.

Target Area
The dataset is a populated area with multiple features. It consists of an assortment of buildings which includes small houses, residential apartments, office complexes, train yards and a convention center. It also features a railway track which encircles the buildings. A moderately busy road goes across the area. There is a green field towards south-west and a forest area towards north-east.
Flight Characteristics

Flight Pattern: Double Lawnmower Grid
Front Overlap: 50%
Side Overlap: 80%
Flight Altitude: 155m / 508 feet AGL
Average GSD: 4.26 cm/pixel / 1.67 in/pixel
Area Covered: 0.223km2 / 55.1235 acres
Inputs
Number of images: 169
Input image resolution: 4608×3456
Camera model name: Canon IXUS 127 HS
Methodology
We will comparing the outputs processed from WebODM against the outputs already present in the demo project.
Processing Options
Pix4D
Pix4D Version: Enterprise 2.1.61
WebODM
ODM version: 2.3.3
dtm: true
dsm: true
gps-accuracy: 0.1
pc-quality: ultra
Result
Quality
On a high-level, the output produced by WebODM looks to be at par with Pix4D. The major differences between the two are most visible in the case of building edges and overground structures. Pix4D produces slightly smoother building edges as compared to WebODM. On the other hand, WebODM is better at handling overground structures such as electric poles and cars which have a ghost-like appearance in Pix4D. Trees are also much more defined in WebODM.
Accuracy
We do comparison for two outputs. One, when GCPs were provided and second, when GCPs were not provided.
With GCP
We have used 5 out of 6 GCPs as GCPs. We use the remaining GCP, GCP 4, as checkpoint to compare accuracy.
Checkpoint | WebODM – X Error | WebODM – Y Error | WebODM – Z Error | Pix4D – X Error | Pix4D – Y Error | Pix4D – Z Error |
GCP 4 | 0.090 | 0.066 | 0.493 | 0.051 | 0.046 | 0.412 |
Though it is not quite able to reach the accuracy provided by Pix4D, WebODM does manage to reach pretty close to be able to useful in many applications.
Without GCP
We can use the GCPs as checkpoints to check accuracy in this case.
Checkpoint | WebODM – X Error | WebODM – Y Error | WebODM – Z Error | Pix4D – X Error | Pix4D – Y Error | Pix4D – Z Error |
GCP 0 | -1.922 | 0.006 | -53.904 | -1.625 | -0.015 | -54.082 |
GCP 1 | -2.398 | -0.536 | -53.530 | -2.410 | -0.507 | -53.849 |
GCP 2 | -1.728 | -0.649 | -54.492 | -1.691 | -0.836 | -54.630 |
GCP 3 | -1.723 | -0.832 | -55.550 | -1.456 | -0.920 | -55.698 |
GCP 4 | -1.582 | -0.241 | -55.083 | -1.280 | -0.290 | -55.430 |
GCP 5 | -2.452 | -0.303 | -53.919 | -2.098 | -0.373 | -54.497 |
RMS | 1.996 | 0.508 | 54.418 | 1.801 | 0.581 | 54.702 |
Both Pix4D and WebODM are well off-the-mark but their divergences are almost equal in magnitude.
Conclusion
In both quality and accuracy, Pix4D demonstrates why it is the preferred solution for thousands of drone pilots. Still, for a free and open-source software that is relatively new, WebODM does hold its ground against the 10-year old proprietary Pix4D. One thing is clear, with another good option available now for photogrammetry, the drone pilots are the real winner.
Managed WebODM Deployment
Due to its open nature of licensing, WebODM allows for powerful cloud deployments which can provide much cheaper and faster processing in an automated fashion than the usual workstation-based deployments. If you have a business requirement for managed WebODM deployments, check out WebODM on Cloud.