Recently, I was tasked with processing nearly 20,000 drone images collected along a road corridor using Agisoft Metashape. It was the largest photogrammetry project I had ever worked on.
To be honest, I wasn’t prepared for it.
I underestimated the scale of the project, overestimated my hardware capabilities, and didn’t fully understand some of the challenges that come with processing large datasets. What followed were long days of troubleshooting, learning new techniques, testing different workflows, and worrying about whether I would be able to deliver.
In the end, I delivered a complete orthomosaic, but not without learning some valuable lessons along the way.
Skill Matters More Than You Think
Before this project, most of the datasets I worked with were relatively straightforward. Images aligned well, processing went smoothly, and the software handled most of the heavy lifting.
This project was different.
Some sections had overlap issues, while others contained large homogeneous surfaces that made it difficult for the software to reconstruct the scene correctly. I had to learn new troubleshooting techniques and gain a deeper understanding of how photogrammetry software actually works rather than relying on default settings.
The experience reminded me that software is only a tool. Understanding the fundamentals is what allows you to solve problems when things go wrong.
Processing Power Is Not Optional
Large drone mapping projects require serious computing resources.
I underestimated how demanding 20,000 images would be. My hardware struggled with processing, forcing me to explore cloud-processing options and alternative workflows. Some of those attempts cost time and money.
One lesson became very clear: before accepting large photogrammetry projects, make sure your hardware matches the scope of the work.
Sometimes the biggest bottleneck isn’t the software—it’s the computer running it.
Understand the Project Before You Accept It
This was probably my biggest mistake.
Had I taken more time to evaluate the dataset size, project requirements, processing complexity, and hardware demands, I would have prepared differently. I might have upgraded my workflow, partnered with someone with stronger computing resources, or even declined the project.
Technical capability is important, but understanding what you’re getting into is equally important.
A large project can quickly become overwhelming if you don’t fully assess the requirements from the start.
Good Outputs Start With Good Flight Planning
One of the most important lessons came from the areas that refused to stitch correctly.
After hours of troubleshooting, it became clear that some sections had gaps in image coverage. The issue wasn’t the processing software—it was the data collection.
This reinforced a lesson every mapping professional eventually learns:
You cannot completely fix poor data collection with software.
Good deliverables start in the field. Proper flight planning, adequate overlap, and consistent image capture are what make successful processing possible. When those elements are missing, fixing the problem often means returning to the site and collecting additional data.
Final Thoughts
This project tested my technical skills, patience, and problem-solving ability more than any previous mapping project.
It also reminded me that growth often comes from taking on challenges that initially feel uncomfortable.
Would I approach a 20,000-image project differently today? Absolutely.
But that’s exactly why the experience was valuable.
Sometimes the most important deliverable isn’t the orthomosaic.
It’s the lessons you take into the next project.