This article explains two possible causes and three solutions when a project fails to process a large dataset.
Error
Failures processing a large dataset.
Description
PIX4Dmapper is able to process unlimited number of images simultaneously. However, depending on the quality of the dataset and the processing resources, there might be some issues with low-quality datasets or datasets larger than 1000 images. This is an indicative value, as processing depends on the image resolution, image content, overlap between images, processing options, and chosen output resolution.
Cause
- Insufficient memory resources: When processing large datasets, the software extracts a very large number of keypoints which are matched between the images to compute 3D points. This amount of points requires a large amount of RAM to be processed simultaneously. The message "Your computer has less than 100 MB RAM available” may appear and the computation may be very slow. In rare cases, processing failure or software instabilities may appear.
- Calibration errors due to heterogeneous datasets: Large projects might have been captured in multiple flights/image acquisition sessions, multiple cameras, or at different dates. This may make the images and camera models of the dataset not homogeneous enough to be processed simultaneously.
a. If the overlap between the datasets is not sufficient, the reconstruction might not be complete.
b. If there are too many changes in the environment between the datasets (sun direction, weather conditions, moving cars, new buildings, snow, etc.), dataset has poor overlap or poor image quality, processing may fail.
Solution
If having insufficient memory resources:
- Increase the RAM of the computer:
Adding more RAM allows the user to have more 3D points simultaneously in memory and therefore to process larger datasets. As an indication, it is recommended to have 64GB of RAM to process projects of more than 1000 images with 24MP having an 80% overlap.
- Generate less 3D points:
For step-by-step instructions: How to change the Processing Options to generate fewer 3D Points.
Warning: Generating less 3D points may lead to lower-quality results.
- Divide the datasets:
Divide the dataset into sub-datasets. For more information: Processing Large Datasets.
If having calibration errors due to a heterogeneous dataset:
- It is needed to acquire the images again. In order to design the flight/image acquisition plan and configure the camera settings: Step 1. Before Starting a Project.
If having calibration errors due to insufficient dataset quality:
- It is needed to acquire the images again. In order to design the flight/image acquisition plan and configure the camera settings: Step 1. Before Starting a Project.