Skip to content

run_pipeline.md

Latest commit

 

History

History
36 lines (31 loc) · 2.54 KB

run_pipeline.md

File metadata and controls

36 lines (31 loc) · 2.54 KB

Run the SSR Pipeline

  • The starting point of the SSR pipeline is ssr/run_pipeline.py

  • The first time ssr/run_pipeline.py is executed, it will create a configuration at ssr/configs/pipeline.toml using the template located at ssr/configs/pipeline_template.toml

  • Adjust the paths in the config file ssr/configs/pipeline.toml

    • For the executables (fp = file path, dp = directory path)

      • colmap_vissat_exe_fp
      • mve_apps_dp
      • texrecon_apps_dp
      • gdmr_bin_dp
      • openmvs_bin_dp
      • colmap_exe_dp

    • For the reconstruction results (i.e. workspace directories)

      • workspace_vissat_dp
      • workspace_ssr_dp
      • meshlab_temp_dp

    • For the input data uncomment the relevant block in the config (depending on the input dataset)

      • Option 1: IARPA MVS3DM dataset
        • dataset_adapter = "ssr.input_adapters.adapter_MVS3DM"
        • satellite_image_pan_dp=/path/to/spacenet-dataset/Hosted-Datasets/MVS_dataset/WV3/PAN
        • satellite_image_msi_dp=/path/to/spacenet-dataset/Hosted-Datasets/MVS_dataset/WV3/MSI
      • Option 2: Data Fusion Contest 2019 (dfc2019) Track 3 dataset
        • dataset_adapter = "ssr.input_adapters.adapter_DFC2019"
        • satellite_image_rgb_tif_dp = "path/to/input/tif/files"
          • Note: the dfc2019 Track 3 dataset contains data for multiple location sites. The SSR pipeline expects satellite_image_rgb_tif_dp to point to a directory containing a subset of the data belonging to the same location (for example JAX_004_*_RGB.tif)
          • If the relevant .txt file from Track 3 / Training data / Reference Train-Track3-Truth.zip (for example JAX_004_DSM.txt ) is placed in the same input directory and ul_easting, ul_northing, width, height are not set, they will be calculated based on the .txt file.

  • Adjust the list of meshing_backends according to the installed MVS / Surface reconstruction libraries

    • For example, comment Openmvs in meshing_backends, if it is not available

  • Run ssr/run_pipeline.py

  • The final output of the pipeline based on the meshing backends chosen can be found in workspace_ssr_dp/surface/surface/mve/
⚡