• Python 3.5, 3.6, 3.7, or 3.8
  • gcc/10.2.0
  • metashape/1.7.3
  • mesa/20.1.6

module load <package>

Agisoft Metashape Tutorial

Step 0: Data Collecting

Camera settings

Any high resolution digital camera (> 5 MPix) can be used for capturing images suitable for 3D model reconstruction in Metashape software. It is suggested to use focal length from 20 to 80 mm interval in 35mm equivalent while avoiding ultra-wide angle and fisheye lenses. Fixed lenses are preferred for more stable results.

Images settings

Take sharp photos at maximal possible resolution with sufficient focal depth and the lowest value of ISO. For the Metashape analysis use RAW data. The lossless conversion to the TIFF format is preferred over JPG, which induce more noise. Also, do not use any pre-processing (resize, rotate, crop, etc.) on your photos.

Optimal image sets

In general, a good set of images is not random. More than required number of photos is better than not enough but redundant or highly overlap pictures are not useful. However, the detail of the geometry should be visible from at least two different camera snapshots. To learn more tips & tricks see the Capturing scenarios and Plan Mission sections in the Metashape user manual, pp 9-14.

Step 1: Loading & Inspecting Photos

Creating application object

import Metashape as MS

doc =

CPU/GPU settings

Metashape exploits GPU processing power that speeds up the process significantly. If you have decided to switch on GPUs to boost the data processing with Metashape, it is recommended to uncheck “Use CPU when performing GPU accelerated processing” option, providing that at least one discrete GPU is utilized for processing. (Preference settings in Metashape User Manual, pp. 15) = 2 ** (len( - 1)         # activate all available GPUs
  if <= 1:                                        # (faster with 1 no difference with 0 GPUs) = True                                    # enable CPU for GPU accelerated processing
  elif > 1:                                       # (faster when multiple GPUs are present) = False                                   # disable CPU for GPU accelerated tasks

Loading images

datadir = "/absolute/path/to/your/input/directory/with/photos"    # directory with image inputs
photo_files = os.listdir(datadir)                                 # list of filenames for photo set
photos = [os.path.join(datadir, p) for p in photo_files]          # convert to full paths

Step 2: Generating Sparse Point Cloud (SPC)

Step 3: Generating Dense Point Cloud (DPC)

Step 4: Generating of a Surface: Mesh or DEM