With the release of Pix4d Matic versions 1.64.0 and 1.65.0, which introduce Native ARM support for Apple Silicon, I wanted to evaluate how these updates impact processing times compared to the latest stable version, 1.63.0. This test was conducted to see if the new features translate into real-world performance gains on Mac OS.

Introduction to Pix4D Matic and Its Relevance for Professional Drone Mapping Services

In the rapidly evolving field of drone mapping services, having the right tools is crucial. At Holland Productions, we use Pix4d Matic to deliver precise and accurate drone mapping solutions for our clients. Whether you're looking for 3D mapping with a drone or professional drone mapping services for construction projects, this software has become an essential part of our toolkit.

As drone professionals, we understand the importance of choosing the right software for commercial drone services. Pix4d Matic excels in processing large datasets quickly, making it ideal for projects requiring photogrammetric data from drones. This is particularly important in sectors like construction and environmental monitoring, where timely and cost-effective drone mapping solutions are crucial.

Our commercial drone services are also highly accurate, partly thanks to Pix4d Matic's seamless integration with ground control points. Using GCPs ensures that our maps and models are not only detailed but also aligned with real-world coordinates, providing the accuracy needed for complex projects. If you're searching for drone services with quick turnaround times, Pix4d Matic helps us meet those demands efficiently.

Importance of Processing Time in Photogrammetry and Its Impact on Drone Mapping Services

Processing time is a critical factor in drone mapping services, especially when clients need results fast. With Pix4d Matic, we can provide drone mapping services with quick turnaround times, ensuring that our clients can make timely decisions based on the most accurate data.

For businesses looking for cost-effective drone mapping solutions, the efficiency of Pix4d Matic allows us to deliver high-quality results without the long wait times typically associated with large-scale photogrammetry projects. Whether it's professional drone mapping services for construction projects or data from drones for environmental monitoring, faster processing means better service for our clients.

Our experience as drone professionals also extends to providing commercial drone services across various industries. By leveraging Pix4d Matic, we can handle everything from drone photography services for marketing to comprehensive drone mapping services for urban planning. This versatility allows us to meet the diverse needs of our clients, no matter the size or complexity of their projects.

Overview of the New Feature: Native ARM Support in Pix4d Matic 1.64.0

Pix4d Matic’s preview version 1.64.0 introduced Native ARM support, optimizing the software for Apple’s M-series chips which is supposed to speed up processing times significantly.

Version 1.65.0, also in preview, builds on these improvements with further optimizations. However, for those prioritizing stability in their drone mapping services, version 1.63.0 remains the latest stable release, ensuring reliable performance for professional use

In this blog, we are going to test these versions on our Mac Studio which is our photo and video editing machine, and has not been used for any photogrammetry processing in Matic due to the long processing times.

Test Setup

To evaluate the performance of Pix4d Matic across different versions, I conducted the tests using the following Mac OS hardware and software environment:

• Hardware:

  • CPU: Apple M1 Ultra, 1 CPU, 20 threads

  • RAM: 64 GB

  • GPU: Apple M1 Ultra (Driver: 4.1 Metal - 88.1)

• Operating System: macOS 14.5

• Storage: The images and the Pix4d Matic project folder were stored on the internal hard drive to ensure optimal data access speeds.

Dataset Details

For consistency, I used the same dataset across all tests:

• Number of Images: 630

• Resolution: 8192x5460 (DJI_ZenmuseP1_24.0)

• Total Data Size: 11.9GB (Images)

Ensuring Consistent Testing Conditions

To ensure that the tests were fair and consistent across all software versions (1.63.0, 1.64.0, and 1.65.0), I took the following steps:

• Identical Hardware Setup: All tests were conducted on the same Mac with no changes to hardware configuration.

• Same Dataset: The exact same dataset was used for each test run to ensure comparability.

• Same Processing Template: The exact same processing template and settings were used for each test. Processing template settings are shown in next section.

• Controlled Environment: The Mac was rebooted before each test to clear memory and ensure that no other applications were running that could affect performance.

• Consistent Storage Location: Both the images and Pix4d Matic project folder were consistently stored on the internal hard drive to avoid any variations in data access speeds.

This setup provided a reliable basis for comparing the processing times and performance across the different versions of Pix4d Matic.

Processing Settings:

Calibration:

• Template: Map

• Pipeline: Trusted location and

• orientation

• Image scale: 1/1

• Internals confidence: Low

• Max extracted keypoints:

• Automatic

• Reoptimized: No

• Use automatic ITPs: Disabled

Dense point cloud

• Algorithm: Hardware

• accelerated

• Image scale: 1/2

• Density: Optimal

• Min number of matches: 3

• Multiscale: Enabled

• Noise filter: Disabled

• Sky filter: Disabled

• Mask-aware: Disabled

Mesh

• Input point cloud: Dense

• Template: Aerial

• Texture size: 8192x8192

• Deghosting: Weak

• Decimation: Limit triangle count

• Maximum triangle count:

• 1000000

• Plane-aware: Disabled

• Sky mask: Disabled

• Smoothing: Enabled

• Mask-aware: Disabled

• Interior improvement: Disabled

DSM

• Input point cloud: Dense

• Interpolation: Enabled

• Resolution: 1.5 cm/px

• Surface smoothing: 12 px

Orthomosaic

• Resolution: 1.5 cm/px

• Algorithm: Hardware

• accelerated

• Oblique: Disabled

• Deghosting: Disabled

Processing Time Tests

Total Processing Time (All 5 Processing Steps):

Calibration Step:

Dense point cloud:

Mesh:

DSM:

Orthomosaic:

Analysis and Discussion

Processing Time Differences Between Versions

The tests revealed significant differences in processing times across the five steps for each version of Pix4d Matic. Here’s a detailed breakdown:

• Calibration:

  • Version 1.63.0: 43 minutes, 14 seconds

  • Version 1.64.0: 29 minutes, 19 seconds

  • Version 1.65.0: 28 minutes, 41 seconds

  • Improvement: Version 1.64.0 reduced the calibration time by approximately 32.2% compared to 1.63.0, with a slight additional improvement in 1.65.0 of 2.2%.

• Dense Point Cloud:

  • Version 1.63.0: 1 hour, 2 minutes, 38 seconds

  • Version 1.64.0: 1 hour, 10 minutes, 27 seconds

  • Version 1.65.0: 1 hour, 9 minutes, 11 seconds

  • Observation: This step took longer in both 1.64.0 and 1.65.0 compared to 1.63.0, with a 12.5% increase in time for 1.64.0, and a slight improvement in 1.65.0, though still slower overall.

• Mesh:

  • Version 1.63.0: 29 minutes, 41 seconds

  • Version 1.64.0: 25 minutes, 16 seconds

  • Version 1.65.0: 16 minutes, 54 seconds

  • Improvement: This step saw a dramatic reduction in time, with 1.64.0 being 14.8% faster than 1.63.0 and 1.65.0 offering a further improvement of 33.1% over 1.64.0.

• DSM:

  • Version 1.63.0: 6 minutes, 18 seconds

  • Version 1.64.0: 4 minutes, 51 seconds

  • Version 1.65.0: 4 minutes, 52 seconds

  • Improvement: Version 1.64.0 reduced the DSM processing time by 23% compared to 1.63.0, with 1.65.0 maintaining this performance.

• Orthomosaic:

  • Version 1.63.0: 2 minutes, 25 seconds

  • Version 1.64.0: 1 minute, 54 seconds

  • Version 1.65.0: 1 minute, 52 seconds

  • Improvement: The orthomosaic generation time improved by 21.5% in 1.64.0 compared to 1.63.0, with a marginal further improvement in 1.65.0.

Total Processing Times

The cumulative effect of these improvements is reflected in the total processing times:

• Version 1.63.0: 2 hours, 24 minutes, 16 seconds

• Version 1.64.0: 2 hours, 11 minutes, 47 seconds

• Version 1.65.0: 2 hours, 1 minute, 30 seconds

These totals show a clear reduction in overall processing time:

• Version 1.64.0 reduced the total processing time by approximately 9.1% compared to 1.63.0.

• Version 1.65.0 further improved total processing time by approximately 7.8% compared to 1.64.0, and by about 15.8% compared to 1.63.0.

Impact of Native ARM Support in Version 1.64.0

The introduction of Native ARM support in version 1.64.0 brought significant performance gains, particularly in calibration and mesh processing. Running Pix4d Matic natively on the Apple M1 Ultra allowed for substantial improvements in efficiency, reducing processing times across most steps.

Overall Performance Improvements from Version 1.63.0 to 1.65.0

From version 1.63.0 to 1.65.0, the enhancements in Pix4d Matic are clear, with total processing time reduced by nearly 16% and individual steps like mesh processing improving by over 40%. These optimizations make the latest preview versions particularly attractive for professionals in drone mapping services who require fast and efficient processing.

Observed Anomalies or Unexpected Results

While most steps saw performance gains, the Dense Point Cloud step exhibited longer processing times in the newer versions. This anomaly might indicate that the new ARM optimizations require further refinement for this specific task. However, the overall gains in speed and efficiency outweigh this issue, making the newer versions, especially 1.65.0, a strong choice for those using Apple Silicon Macs.

Conclusion

The tests on Pix4d Matic versions 1.63.0, 1.64.0, and 1.65.0 reveal clear advantages in upgrading to the newer versions, especially for users on Apple Silicon. The introduction of Native ARM support in version 1.64.0 led to significant reductions in total processing time, which dropped by 9.1% compared to the stable 1.63.0. Version 1.65.0 further improved this by an additional 7.8%, resulting in an overall reduction of nearly 16% from the original baseline.

For professionals offering drone mapping services, these improvements translate into faster project turnaround and increased capacity to handle more complex tasks. Despite the slight increase in processing time for the Dense Point Cloud step, the substantial gains in other areas make the latest versions, particularly 1.65.0, a valuable upgrade.

In conclusion, whether you prioritize stability with version 1.63.0 or seek the cutting-edge performance of the preview versions, Pix4d Matic remains an essential tool for delivering high-quality, efficient drone mapping services.