Traditionally, there have been five cost models when it came to drone mapping software:
1. Pay one-time for a perpetual desktop-bound license, ex. perpetual licenses for Pix4D Mapper, Agisoft Metashape
2. Pay one-time for a perpetual license and deploy it in cloud, ex. Pix4D Mapper Enterprise Server License, Agisoft Metashape
3. Pay a subscription fee for a desktop-bound license, ex. monthly subscription for Pix4D Mapper
4. Pay a subscription fee for a cloud-based processing service, ex. Pix4D Cloud, DroneDeploy
5. Pay a pay-per-use fee for a cloud-based processing service, ex. Maps Made Easy
But now things are changing with the advent of an open-source solution, WebODM. For those who don’t know, WebODM is a free and open-source drone mapping software. It is able to generate photogrammetric outputs on par with the other softwares mentioned above. Therefore there is now one more cost model available to us:
6. You deploy WebODM on a cloud service such as AWS or Google Cloud and process your datasets on the cloud.
All these cost models have their own implications on number of images that can be processed, time to process and how much it will cost. In order to analyze these parameters we are going to take a test scenario.
Test Scenario
Let’s say our monthly surveyed area is 25,000 hectare (250 sq.km.). It will generate around 50,000 images (assume a DJI Camera, 4000×3000 image resolution, 400 feet flight altitude with 60% overlap and 2 inch/pixel GSD). Let’s assume we get 20 sets of images, each set consisting of 2,500 images. We will now calculate the cost to process this dataset to produce orthomosaics in different software setups.
Requirements
A minimum RAM of 64GB will be required on the desktop or a cloud server in order to process 1 set of 2,500 images. We’ll take this Dell XPS CPU as a reference to set our cost of acquiring a desktop with 64GB RAM at $2,400.
Assumptions
- Although processing time will vary with each software and the specified parameters, for ease of calculation, we will assume that it will take 48 hours to process 2,500 images on a desktop with 64GB RAM and an appropriate number of cores.
- On 1 desktop, the batches can only be processed serially, i.e., one after the other.
Ignored Costs
We will ignore the following costs in our analysis:
- Electricity: This is completely dependent on the location so we will ignore this.
- Cloud storage and data transfer: Storage and data transfer is quite cheap on the cloud and would not make a significant difference to our calculations.
Comparison
Cost Model 1: Perpetual License on Desktop
We would need 2 desktops to keep the monthly processing time under 30 days.
Pix4D Mapper
Pix4D Mapper’s Perpetual License comes at $4,990. We would need 2 licenses, 1 for each desktop.
Hardware Cost: $2,400 x 2 = $4,800
License Cost: $4,990 x 2 = $9,980
Processing Time: 48 x 20 / 2 = 480 hours, or 20 days
Agisoft Metashape
We will use Professional Edition of Agisoft Metashape instead of the Standard Edition since Standard Edition does not support GCPs, DSM/DTM etc. License cost for Professional Edition is $3,499.
Hardware Cost: $2,400 x 2 = $4,800
License Cost: $3,499 x 2 = $6,998
Processing Time: 48 x 20 / 2 = 480 hours, or 20 days
WebODM
Hardware Cost: $2,400 x 2 = $4,800
License Cost: $0
Monthly Processing Time: 48 x 20 / 2= 480 hours, or 20 days
Cost Model 2: Perpetual License on Cloud
Pix4D Mapper
Pix4D requires the users to purchase an Enterprise Server License to be able to run Pix4D in the cloud. The pricing for Enterprise Server License is not disclosed publicly by Pix4D. Therefore we will not include Pix4D Cloud Deployment in our calculations.
Agisoft Metashape
Agisoft Metashape allows running the Node-locked License on a cloud server provided a few conditions are met. We will use the r5.2xlarge AWS EC2 On-demand instance as reference for the server. We will assume 2 servers are run full-time, since the license activation has to be maintained.
License Cost: $3,499 x 2 = $6,998
Monthly Server Operating Cost (Processing Cost): $0.504 * 24 * 30 * 2 = $725.76
Processing Time: 48 x 20 / 2 = 480 hours, or 20 days
Cost Model 3: Subscription License for 2 Desktops
Pix4D Mapper
Yearly subscription fee for Pix4D Mapper is $3,500.
Hardware Cost: $2,400 x 2 = $4,800
Yearly Subscription Cost: $3,500 x 2 = $7,000
Processing Time: 48 x 20 / 2 = 480 hours, or 20 days
Cost Model 4: Subscription of Cloud-based Processing Service
Pix4D Cloud
Cannot be used. Large Plan has a maximum limit of 10,000 images per month.
DroneDeploy
We could not find any publicly available information about limit on the number of projects or images on DroneDeploy’s website but we have been told by our readers that the usage limit on Business plan is 25 maps or 8500 images per month. Thus we cannot use DroneDeploy Cloud Subscription as well.
WebODM Lightning
WebODM Lightning is a cost-effective subscription-based cloud processing service by the developers of WebODM. Business Plan has a monthly subscription cost of $99 and allows processing 3000 images/map with max 4 concurrent tasks.
Yearly Subscription Cost: $1,188
Processing Time: 48 x 20 / 4 = 240 hours, or 10 days
Cost Model 5: Pay-per-use Cloud-based Processing Service
Maps Made Easy
According to their pricing chart, it would cost $669.99 to process images for an area of about 121 sq. km.
Processing Cost for 250 sq km: $1,384
Processing Time: 2 days
WebODM Lightning
WebODM Lightning also offers a credit-based pricing model where it charges $99.97 for 57.48 sq. km.
Processing Cost for 250 sq km: $434
Processing Time: 2 days
Cost Model 6: Deploying WebODM on Cloud Service
In this model, you will need to pay costs for running servers on cloud, data storage, data transfer and few other minor things (such as Public IP address). But by far the major cost will be of the processing servers being used to run the tasks. Therefore we will only focus on the processing server costs.
AWS (Amazon Web Services)
We will take AWS as reference for doing our cost calculations. Server units in AWS are provided in the form of EC2 instances. There are many types of EC2 instances available with different configurations of memory, number of cores and type of processors. For this analysis, let us go ahead with an r5.2xlarge instance. An r5.2xlarge instance has the following parameters:
- vCPU: 8
- Memory (GiB): 64
Each vCPU is a thread of a CPU core. See AWS documentation for more details on vCPU and other instance details.
With AWS, we can provision multiple instances simultaneously. This will allow us to run all the batches in parallel. Also, we will be able to shut down these instances when our tasks are done which will result in significant cost-savings.
AWS provides multiple ways to pay for an EC2 instance. These are On-demand, Reserved, Spot Instances, Savings Plans and Dedicated Hosts. We’ll not go into the details of these here but Spot Instances are very cost-effective as short-lived instances so we will use Spot Instance pricing for our calculations.
Processing Cost: $81.6
Processing Time: 2 Days
Visual Comparison
Deploying WebODM on Cloud
Unless you are an expert with cloud deployments, it can be a bit tricky to deploy WebODM on cloud and optimize it for maximum cost-savings. If you need expert assistance in deploying WebODM on cloud, reach out to webodm@algopixel.tech.
