The tables in this article shows the memory consumption peaks and spent time for the common processing operations for different types of surveys: nadir aerial survey and close-range survey in order to give the general understanding of the time and RAM required for regular workflow procedure.


The processing has been performed on Agisoft Cloud nodes in Metashape Professional with the following hardware configuration (using single node for each operation):


CPU32 vCPU (2.7 GHz Intel Xeon E5 2686 v4)
GPU2 x NVIDIA Tesla M60
RAM240 GB



Aerial data processing benchmark



The dataset used to benchmark aerial data processing consists of 920 images, 40 MPix resolution each. The data has been acquired using Wingtra platform (WingtraOne VTOL mapping drone) with Sony RX1RII camera, coordinates measured with a precise PPK method. No GCPs data was used for processing. For the evaluation purposes the dataset can be downloaded from the Wingtra website ("High-resolution quarry mapping" dataset): 

https://wingtra.com/mapping-drone-wingtraone/aerial-map-types/#3d-reconstruction



The alignment has been performed using Medium and High accuracy in turns with Generic+Reference preselection and 40,000 / 4,000 key/tie point limits for both cases. Adaptive camera model fitting - disabled in each processing scenario.


All further processing operations have been performed based on the alignment corresponding to High accuracy matching.

Image matching, Depth maps calculation and depth maps based mesh and DEM generation operations used GPU acceleration.


Table 1. Aerial data processing benchmark in Metashape Professional 1.7.2:


Task (Source)/ ParametersTimeMemory Peak
Match Photos
Medium accuracy23 m 02 s758 Mb
High accuracy22 m 46 s603 Mb
Align cameras
(Medium accuracy)7 m 27 s416 Mb
(High accuracy)7 m 08 s427 Mb
Depth maps
High/Mild3 h 08 m20.41 Gb
Medium/Mild     47 m  8.51 Gb
Low/Mild19 m 23 s  5.99 Gb
Dense Cloud
High/Mild2 h 36 m15.03 Gb
Medium/Mild      37 m13.28 Gb
Low/Mild12 m 33 s  7.51 Gb
Mesh (from Depth Maps)
High/Mild10 h 12 m14.80 Gb
Medium/Mild  2 h 15 m13.04 Gb
Low/Mild       43 m 10.80 Gb
DEM (from Dense Cloud)
High16 m 28 s350 Mb
Medium  5 m 335 Mb
Low  2 m331 Mb
DEM (from Depth Maps)
High3 h 37 m12.18 Gb
Medium1 h 06 m  8.13 Gb
Low
25 m 26 s  5.59 Gb
Orthomosaic (DEM)
DEM (High)3 h 05 m  4.89 Gb
DEM (Medium)1 h 30 m  8.96 Gb
DEM (Low)1 h 16 m11.94 Gb
Tiled Model (from Depth maps, texture on GPU)
High  43 h15.05  Gb
Medium  9 h 19 m10.49 Gb
Low  4 h 38 m  6.81 Gb


Memory consumption for the orthorectification process (the first part of the Build Orthomosaic stage) is proportional to the number of CPU threads.




Close-range data processing benchmark




The dataset used for close-range processing performance benchmarking has been captured with DJI Phantom 4 RTK drone and consists of 124 overhead flight images (nadir and oblique) and 648 images taken by the automated mission plan following. In total 772 images of 18 MPix were used.


The overview of the reconstructed mesh model with the camera locations and working volume is shown on the following screenshot:




The alignment has been performed using Medium and High accuracy in turns with Generic preselection only and 40,000 / 4,000 key/tie point limits for both cases. Adaptive camera model fitting and Exclude stationary points options - disabled in each processing scenario.


All further processing operations have been performed based on the alignment corresponding to High accuracy matching.


Image matching, Depth maps calculation and depth maps based mesh generation operations used GPU acceleration. 



Table 2. Close-range data processing benchmark in Metashape Professional 1.7.2:


Task (Source) / ParametersTimeMemory
Peak
Match Photos
Medium accuracy30 m 43 s466 Mb
High accuracy36 m 21 s522 Mb
Align cameras
(Medium accuracy)8 m 24 s785 Mb
(High accuracy)10 m 27 s880 Mb
Depth maps
High/Mild43 m8.02 Gb
Medium/Mild14 m3.72 Gb
Low/Mild7 m2.38 Gb
Dense Cloud
High/Mild1 h 24 m25.88 Gb
Medium/Mild24 m 13 s14.94 Gb
Low/Mild10 m 31 s  5.34 Gb
Mesh (from Depth Maps)
High/Mild2 h 59 m11.22 Gb
Medium/Mild1 h 8 m8.89 Gb
Low/Mild26 m 33 s9.16 Gb
Texture UV (Generic mapping)
High (32.2 M faces)51 m7.28 Gb
Medium (8.9 M faces)33 m6.12 Gb
Low (2.4 M faces)22 m 6.78 Gb
Texture Blending (on GPU only, 16K)

Medium (8.9 M faces )
22 m 50 s10.94 Gb
6 Gb VRAM
Low (2.4 M faces )
14 m 44 s10.05 Gb
6 Gb VRAM
Texture Blending (on CPU only, 16K)
High (32.2 M faces)48 m33.86 Gb
Medium (8.9 M faces)24 m32.95 Gb
Low (2.4 M faces)17 m 32.86 Gb
Tiled Model (from Depth Maps, texture on CPU)
High/Mild
7 h 43 m27.86 Gb
Medium/Mild
4 h 8 m28.14 Gb
Low/Mild
3 h 22 m26.47 Gb

 

Memory consumption for the texture blending operation (the second part of the Build Texture stage) is proportional to the number of CPU threads, if performed on CPU. There's no such limitation when texture blending uses GPU acceleration.