.github/workflows/publish-windows.yml

@@ -9,7 +9,7 @@ on:
 
 jobs:
   build:
-    runs-on: windows-2022
+    runs-on: windows-2019
     steps:
     - name: Checkout
       uses: actions/checkout@v2

.gitignore

@@ -35,4 +35,3 @@ venv/
 python38/
 dist/
 innosetup/
-.DS_Store

README.md

@@ -26,8 +26,7 @@ Run ODM by placing some images (JPEGs, TIFFs or DNGs) in a folder named “image
 ```bash
 # Windows
 docker run -ti --rm -v c:/Users/youruser/datasets:/datasets opendronemap/odm --project-path /datasets project
-```
+
-```bash
 # Mac/Linux
 docker run -ti --rm -v /home/youruser/datasets:/datasets opendronemap/odm --project-path /datasets project
 ```
@@ -95,7 +94,7 @@ run C:\Users\youruser\datasets\project  [--additional --parameters --here]
 ODM has support for doing SIFT feature extraction on a GPU, which is about 2x faster than the CPU on a typical consumer laptop. To use this feature, you need to use the `opendronemap/odm:gpu` docker image instead of `opendronemap/odm` and you need to pass the `--gpus all` flag:
 
 ```
-docker run -ti --rm -v c:/Users/youruser/datasets:/datasets --gpus all opendronemap/odm:gpu --project-path /datasets project --feature-type sift
+docker run -ti --rm -v c:/Users/youruser/datasets:/datasets --gpus all opendronemap/odm:gpu --project-path /datasets project
 ```
 
 When you run ODM, if the GPU is recognized, in the first few lines of output you should see:

SuperBuild/CMakeLists.txt

@@ -184,7 +184,7 @@ set(custom_libs OpenSfM
 
 externalproject_add(mve
     GIT_REPOSITORY  https://github.com/OpenDroneMap/mve.git
-    GIT_TAG         356
+    GIT_TAG         290
     UPDATE_COMMAND  ""
     SOURCE_DIR      ${SB_SOURCE_DIR}/mve
     CMAKE_ARGS      ${WIN32_CMAKE_ARGS} ${APPLE_CMAKE_ARGS}

VERSION

@@ -1 +1 @@
-3.5.6
+3.5.5

configure.sh

@@ -58,7 +58,7 @@ ensure_prereqs() {
     if [[ "$UBUNTU_VERSION" == *"20.04"* ]]; then
         echo "Enabling PPA for Ubuntu GIS"
         sudo $APT_GET install -y -qq --no-install-recommends software-properties-common
-        sudo add-apt-repository ppa:ubuntugis/ppa
+        sudo add-apt-repository -y ppa:ubuntugis/ubuntugis-unstable
         sudo $APT_GET update
     fi
 

contrib/fix_ply/README.md

@@ -1,19 +0,0 @@
-# Fix Ply
-
-Use to translate a modified ply into a compatible format for subsequent steps in ODM. Via Jaime Chacoff, https://community.opendronemap.org/t/edited-point-cloud-with-cloudcompare-wont-rerun-from-odm-meshing/21449/6
-
-The basic idea is to process through ODM until the point cloud is created, use a 3rd party tool, like CloudCompare to edit the point cloud, and then continue processing in OpenDroneMap.
-
-This useful bit of python will convert the PLY exported from CloudCompare back into a compatible format for continued processing in OpenDroneMap.
-
-1. Run project in WebODM and add this to your settings: `end-with: odm-filterpoints`
-1. Once complete, go to your NodeODM container and copy `/var/www/data/[Task ID]/odm-filterpoints` directory
-1. Open CloudCompare and from `odm-filterpoints` directory you've copied, open `point_cloud.ply`
-1. In the box that pops up, add a scalar field `vertex - views`
-1. To see the actual colours again - select the point cloud, then in properties change colours from "Scalar field" to "RGB"
-1. Make your changes to the point cloud
-1. Compute normals (Edit > Normals > Compute)
-1. Save PLY file as ASCII
-1. Run Python file above to fix PLY file and convert to binary
-1. Copy `odm_filterpoints` directory (or just `point_cloud.ply`) back into NodeODM container
-1. Restart project in WebODM "From Meshing" (don't forget to edit settings to remove `end-with: odm-filterpoints` or it's not going to do anything).

contrib/fix_ply/fix_ply.py

@@ -1,68 +0,0 @@
-import os
-import logging
-from plyfile import PlyData, PlyElement
-import numpy as np
-
-# Configure logging
-logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
-
-def pcd_ascii_to_binary_ply(ply_file: str, binary_ply: str) -> None:
-    """Converts ASCII PLY to binary, ensuring 'views' is present and of type uchar.
-       Raises ValueError if neither 'scalar_views' nor 'views' is found.
-    """
-
-    try:
-        logging.info(f"Reading ASCII PLY file: {ply_file}")
-        ply_data: PlyData = PlyData.read(ply_file)
-    except FileNotFoundError:
-        logging.error(f"File not found: {ply_file}")
-        return
-    except Exception as e:
-        logging.error(f"Error reading PLY file: {e}")
-        return
-
-    new_elements: list[PlyElement] = []
-
-    for element in ply_data.elements:
-        new_data = element.data.copy()
-
-        if 'scalar_views' in element.data.dtype.names:
-            new_data['views'] = new_data['scalar_views'].astype('u1')
-            del new_data['scalar_views']
-        elif 'views' in element.data.dtype.names:
-            new_data['views'] = new_data['views'].astype('u1')
-        else:
-            raise ValueError(f"Neither 'scalar_views' nor 'views' found - did you import them when opened the file in CloudCompare?")
-
-
-        new_element = PlyElement.describe(new_data, element.name)
-        new_elements.append(new_element)
-
-    new_ply_data = PlyData(new_elements, text=False)
-
-    try:
-        logging.info(f"Writing binary PLY file: {binary_ply}")
-        new_ply_data.write(binary_ply)
-    except Exception as e:
-        logging.error(f"Error writing PLY file: {e}")
-        return
-
-    logging.info("PLY conversion complete.")
-
-
-if __name__ == '__main__':
-
-    # Parameters
-    base: str = os.path.dirname(os.path.abspath(__file__))
-    ply_file: str = os.path.join(base, 'point_cloud_ascii.ply')
-    binary_ply_file: str = os.path.join(base, 'point_cloud.ply')
-
-    if not os.path.exists(ply_file):
-        logging.error(f"Input file not found: {ply_file}")
-        exit(1) # Exit with error code
-
-    try:
-        pcd_ascii_to_binary_ply(ply_file, binary_ply_file)
-    except ValueError as e:
-        logging.error(f"PLY conversion failed: {e}")
-        exit(1) # Exit with error code to indicate failure

contrib/time-sift/README.md

@@ -1,64 +0,0 @@
-# Plugin Time-SIFT
-
-This script does Time-SIFT processing with ODM. Time-SIFT is a method for multi-temporal analysis without the need to co-registrate the data.
-
->  D. Feurer, F. Vinatier, Joining multi-epoch archival aerial images in a single SfM block allows 3-D change detection with almost exclusively image information, ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146, 2018, Pages 495-506, ISSN 0924-2716, doi: 10.1016/j.isprsjprs.2018.10.016
-(https://doi.org/10.1016/j.isprsjprs.2018.10.016)
-
-## Requirements
-* ODM ! :-) 
-* subprocess
-* json
-* os
-* shutil
-* pathlib
-* sys
-* argparse
-* textwrap
-
-## Usage
-
-### Provided example
-Download or clone [this repo](https://forge.inrae.fr/Denis.Feurer/timesift-odm-data-example.git) to get example data.
-
-Then execute
-```
-python Timesift_odm.py datasets --end-with odm_filterpoints
-```
-It should make the Time-SIFT processing on the downloaded example data, stopping after the filtered dense clouds step.
-
-In the destination dir, you should obtain new directories, ```0_before``` and ```1_after``` at the same level as the ```time-sift-block``` directory. These new directories contain all the results natively co-registered.
-
-You can then use [CloudCompare](https://cloudcompare.org/) to compute distance between the ```datasets/0_before/odm_filterpoints/point_cloud.ply``` and the ```datasets/1_after/odm_filterpoints/point_cloud.ply``` and obtain this image showing the difference between the two 3D surfaces. Here, two soil samples were excavated as can be seen on the image below.
-![](https://forge.inrae.fr/Denis.Feurer/timesift-odm-data-example/-/raw/main/Example.png?ref_type=heads)
-
-### Your own data
-In your dataset directory (usually ```datasets```, but you can have chosen another name) you have to prepare a Time-SIFT project directory (default name : ```time-sift-block```, *can be tuned via a parameter*) that contains :
-   * ```images/``` : a subdirectory with all images of all epochs. This directory name is fixed as it is the one expected by ODM
-   * ```images_epochs.txt``` : a file that has the same format as the file used for the split and merge ODM function. This file name *can be tuned via a parameter*.
-
-The ```images_epochs.txt``` file has two columns, the first column contains image names and the second contains the epoch name as follows
-```
-DSC_0368.JPG 0_before
-DSC_0369.JPG 0_before
-DSC_0370.JPG 0_before
-DSC_0389.JPG 1_after
-DSC_0390.JPG 1_after
-DSC_0391.JPG 1_after
-```
-
-Your directory, before running the script, should look like this :
-```
-$PWD/datasets/
-└── time-sift-block/
-    ├── images/
-    └── images_epochs.txt
-```
-
-At the end of the script you obtain a directory by epoch (at the same level as the Time-SIFT project directory). Each directory is processed with images of each epoch and all results are natively co-registered due to the initial sfm step done with all images.
-```
-$PWD/datasets/
-├── 0_before/
-├── 1_after/
-└── time-sift-block/
-```

contrib/time-sift/Timesift_odm.py

@@ -1,167 +0,0 @@
-# Script for Time-SIFT multi-temporal images alignment with ODM
-#
-# This is python script for ODM, based on the following publication :
-#
-# D. Feurer, F. Vinatier, Joining multi-epoch archival aerial images in a single SfM block allows 3-D change detection
-#    with almost exclusively image information, ISPRS Journal of Photogrammetry and Remote Sensing, Volume 146, 2018,
-#    Pages 495-506, ISSN 0924-2716, https://doi.org/10.1016/j.isprsjprs.2018.10.016.
-
-import subprocess
-import json
-import os
-import shutil
-from pathlib import Path
-import sys
-import argparse
-import textwrap
-
-def main(argv):
-    # Parsing and checking args
-    parser = argparse.ArgumentParser(
-        formatter_class=argparse.RawDescriptionHelpFormatter, description=textwrap.dedent('''\
-            Timesift_odm.py datasetdir [-t <timesift-dir>] [-i <imageepochs-file>] [<options passed to ODM>]
-
-                you can add options passed to ODM, for instance [--end-with odm_filterpoints] so that the final step is point clouds
-                these options are not checked for before the final runs of each epoch, so use it carefully
-            '''))
-    parser.add_argument('datasetdir', help='dataset directory')
-    parser.add_argument('-t', '--timesift-dir',
-                        help='Time-SIFT directory ; default value : "time-sift-block" # must be in the datasetdir')
-    parser.add_argument('-i', '--imageepochs-file',
-                        help='Text file describing epochs ; default value : "images_epochs.txt" # must be in the TIMESIFT_DIR ')
-    args, additional_options_to_rerun = parser.parse_known_args()
-    datasets_DIR = Path(args.datasetdir).absolute().as_posix()
-    if args.timesift_dir:
-        timesift_DIR = args.timesift_dir
-    else:
-        timesift_DIR = 'time-sift-block'
-    if args.imageepochs_file:
-        images_epochs_file = args.imageepochs_file
-    else:
-        images_epochs_file = 'images_epochs.txt'
-    if '-h' in sys.argv or '--help' in sys.argv:
-        parser.print_help()
-        sys.exit()
-    if additional_options_to_rerun:  # for instance, --end-with odm_filterpoints
-        print(f'[Time-SIFT]    Options passed to ODM for the final steps: {additional_options_to_rerun}')
-        print(f'[Time-SIFT]    \033[93mWARNING there is no check of these options done before the last ODM call\033[0m')
-
-    def check_path_args(var: Path):
-        if not var.exists():
-            print(
-                f'\033[91m[Time-SIFT]    ERROR: the {var.as_posix()} directory does not exist. Exiting program\033[0m')
-            exit()
-
-    check_path_args(Path(datasets_DIR))
-    check_path_args(Path(datasets_DIR, timesift_DIR))
-    check_path_args(Path(datasets_DIR, timesift_DIR, images_epochs_file))
-
-    def clean_reconstruction_dict(subdict, key, images):
-        """
-        Delete subdict elements where the key do not match any name in the images list.
-        To create the {epoch} block with only images of this epoch
-        """
-        # The list of valid images is prepared by removing any extension (to be robust to the .tif added by ODM)
-        valid_images = {os.path.basename(image).split(os.extsep)[0] for image in images}
-        for item_key in list(subdict[key]):
-            image_name = os.path.basename(item_key).split(os.extsep)[0]
-            if image_name not in valid_images:
-                del subdict[key][item_key]
-
-    ### Read images.txt file and create a dict of images/epochs
-    images_epochs_dict = {}
-    with open(Path(datasets_DIR, timesift_DIR, images_epochs_file), 'r') as f:
-        for line in f:
-            line = line.strip()
-            if not line:
-                continue  # Empty lines are skipped
-            image, epoch = line.split()
-            if epoch not in images_epochs_dict:
-                images_epochs_dict[epoch] = []
-            images_epochs_dict[epoch].append(image)
-
-    ### Check for existing epochs directories before computing anything (these directories must be deleted by hand)
-    path_exists_error = False
-    for epoch in images_epochs_dict:
-        if Path(datasets_DIR, epoch).exists():
-            if path_exists_error:
-                print(f"sudo rm -rf {Path(datasets_DIR, epoch).as_posix()}")
-            else:
-                print(f'\033[91m[Time-SIFT]    ERROR: {Path(datasets_DIR, epoch).as_posix()} already exists.\033[0m')
-                print(f"               Other epochs probably also exist.")
-                print(
-                    f"               The problem is \033[93mI CAN'T\033[0m delete it by myself, it requires root privileges.")
-                print(
-                    f"               The good news is \033[92mYOU CAN\033[0m do it with the following command (be careful).")
-                print(f'\033[91m               => Consider doing it (at your own risks). Exiting program\033[0m')
-                print(f"- Commands to copy/paste (I'm kind, I prepared all the necessary commands for you).")
-                print(f"sudo rm -rf {Path(datasets_DIR, epoch).as_posix()}")
-                path_exists_error = True
-    if path_exists_error:
-        exit()
-
-    ### LAUNCH global alignment (Time-SIFT multitemporal block)
-    try:
-        subprocess.run(['docker', 'run', '-i', '--rm', '-v', datasets_DIR + ':/datasets',
-                        'opendronemap/odm', '--project-path', '/datasets', timesift_DIR, '--end-with', 'opensfm'])
-    except:
-        print(f'\033[91m[Time-SIFT]    ERROR: {sys.exc_info()[0]}\033[0m')
-        exit()
-    print('\033[92m[Time-SIFT]    Sfm on multi-temporal block done\033[0m')
-
-    print('[Time-SIFT]    Going to dense matching on all epochs...')
-    ### Loop on epochs for the dense matching
-    for epoch in images_epochs_dict:
-        #### We first duplicate the time-sift multitemporal block to save sfm results
-        shutil.copytree(Path(datasets_DIR, timesift_DIR),
-                        Path(datasets_DIR, epoch))
-
-        #### Reads the datasets/{epoch}/opensfm/undistorted/reconstruction.json file that has to be modified
-        with open(Path(datasets_DIR, epoch, 'opensfm', 'undistorted', 'reconstruction.json'), mode="r",
-                  encoding="utf-8") as read_file:
-            reconstruction_dict = json.load(read_file)
-
-        #### Removes images in this json dict (we delete the shot and the rig_instances that do not correspond to this epoch)
-        images = images_epochs_dict[epoch]
-        clean_reconstruction_dict(reconstruction_dict[0], 'shots', images)
-        clean_reconstruction_dict(reconstruction_dict[0], 'rig_instances', images)
-
-        #### Makes a backup of the reconstruction.json file and writes the modified json
-        shutil.copy(Path(datasets_DIR, epoch, 'opensfm', 'undistorted', 'reconstruction.json'),
-                    Path(datasets_DIR, epoch, 'opensfm', 'undistorted', 'reconstruction.json.bak'))
-        with open(Path(datasets_DIR, epoch, 'opensfm', 'undistorted', 'reconstruction.json'), mode="w",
-                  encoding="utf-8") as write_file:
-            json.dump(reconstruction_dict, write_file)
-
-        #### Launches dense matching from the good previous step, with possible options (e.g. => to stop at the point clouds)
-        command_rerun = ['docker', 'run', '-i', '--rm', '-v', datasets_DIR + ':/datasets',
-                             'opendronemap/odm',
-                             '--project-path', '/datasets', epoch,
-                             '--rerun-from', 'openmvs']
-        if additional_options_to_rerun:
-            print(f'[Time-SIFT]        Epoch {epoch}: Rerun with additionnal options: {additional_options_to_rerun}')
-            command_rerun.extend(additional_options_to_rerun)
-        else:
-            print(f'[Time-SIFT]        Epoch {epoch}: Default full rerun')
-        result = subprocess.run(command_rerun)
-        if result.returncode != 0:
-            print(f'\033[91m[Time-SIFT]    ERROR in processing epoch {epoch}\033[0m')
-            print(f'{result=}')
-            exit(result.returncode)
-        print(f'\033[92m[Time-SIFT]        Epoch {epoch} finished\033[0m')
-
-    print('§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§')
-    print('§§§      §§  §§§  §§§§§  §§§      §§§§§§§§§§      §§  §§      §§      §§§')
-    print('§§§§§  §§§§  §§§   §§§   §§§  §§§§§§§§§§§§§   §§§§§§  §§  §§§§§§§§  §§§§§')
-    print('§§§§§  §§§§  §§§    §    §§§    §§§§    §§§§§   §§§§  §§    §§§§§§  §§§§§')
-    print('§§§§§  §§§§  §§§  §§§§§  §§§  §§§§§§§§§§§§§§§§§   §§  §§  §§§§§§§§  §§§§§')
-    print('§§§§§  §§§§  §§§  §§§§§  §§§      §§§§§§§§§      §§§  §§  §§§§§§§§  §§§§§')
-    print('§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§§')
-    print('  \033[92mTime-SIFT with ODM finished, congrats !\033[0m Want to cite the method ?')
-    print('=> D. Feurer, F. Vinatier, Joining multi-epoch archival aerial images in ')
-    print('   a single SfM block allows 3-D change detection with almost exclusively')
-    print('   image information, ISPRS Journal of Photogrammetry and Remote Sensing,')
-    print('   2018, https://doi.org/10.1016/j.isprsjprs.2018.10.016                ')
-
-if __name__ == "__main__":
-    main(sys.argv[1:])

gpu.Dockerfile

@@ -1,8 +1,8 @@
-FROM nvidia/cuda:11.8.0-devel-ubuntu20.04 AS builder
+FROM nvidia/cuda:11.2.2-devel-ubuntu20.04 AS builder
 
 # Env variables
 ENV DEBIAN_FRONTEND=noninteractive \
-    PYTHONPATH="/code/SuperBuild/install/lib/python3.9/dist-packages:/code/SuperBuild/install/lib/python3.8/dist-packages:/code/SuperBuild/install/bin/opensfm" \
+    PYTHONPATH="$PYTHONPATH:/code/SuperBuild/install/lib/python3.9/dist-packages:/code/SuperBuild/install/lib/python3.8/dist-packages:/code/SuperBuild/install/bin/opensfm" \
     LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/code/SuperBuild/install/lib"
 
 # Prepare directories
@@ -21,11 +21,12 @@ RUN bash configure.sh clean
 
 ### Use a second image for the final asset to reduce the number and
 # size of the layers.
-FROM nvidia/cuda:11.8.0-runtime-ubuntu20.04
+FROM nvidia/cuda:11.2.2-runtime-ubuntu20.04
+#FROM nvidia/cuda:11.2.0-devel-ubuntu20.04
 
 # Env variables
 ENV DEBIAN_FRONTEND=noninteractive \
-    PYTHONPATH="/code/SuperBuild/install/lib/python3.9/dist-packages:/code/SuperBuild/install/lib/python3.8/dist-packages:/code/SuperBuild/install/bin/opensfm" \
+    PYTHONPATH="$PYTHONPATH:/code/SuperBuild/install/lib/python3.9/dist-packages:/code/SuperBuild/install/lib/python3.8/dist-packages:/code/SuperBuild/install/bin/opensfm" \
     LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/code/SuperBuild/install/lib" \
     PDAL_DRIVER_PATH="/code/SuperBuild/install/bin"
 
@@ -37,7 +38,7 @@ COPY --from=builder /code /code
 # Copy the Python libraries installed via pip from the builder
 COPY --from=builder /usr/local /usr/local
 #COPY --from=builder /usr/lib/x86_64-linux-gnu/libavcodec.so.58 /usr/lib/x86_64-linux-gnu/libavcodec.so.58
-
+RUN dpkg --remove cuda-compat-11-2
 RUN apt-get update -y \
  && apt-get install -y ffmpeg libtbb2
 # Install shared libraries that we depend on via APT, but *not*

opendm/ai.py

@@ -5,7 +5,6 @@ import zipfile
 import time
 import sys
 import rawpy
-import cv2
 
 def read_image(img_path):
     if img_path[-4:].lower() in [".dng", ".raw", ".nef"]:

opendm/config.py

@@ -41,7 +41,6 @@ rerun_stages = {
     'geo': 'dataset',
     'gltf': 'mvs_texturing',
     'gps_accuracy': 'dataset',
-    'gps_z_offset': 'dataset',
     'help': None,
     'ignore_gsd': 'opensfm',
     'matcher_neighbors': 'opensfm',
@@ -786,12 +785,11 @@ def config(argv=None, parser=None):
                         action=StoreValue,
                         metavar='<positive integer>',
                         default=150,
-                        help='Radius of the overlap between submodels in meters. '
+                        help='Radius of the overlap between submodels. '
                         'After grouping images into clusters, images '
                         'that are closer than this radius to a cluster '
                         'are added to the cluster. This is done to ensure '
-                        'that neighboring submodels overlap. All images' \
+                        'that neighboring submodels overlap. Default: %(default)s')
-                        'need GPS information. Default: %(default)s')
 
     parser.add_argument('--split-image-groups',
                         metavar='<path string>',
@@ -847,16 +845,6 @@ def config(argv=None, parser=None):
                         'set accordingly. You can use this option to manually set it in case the reconstruction '
                         'fails. Lowering this option can sometimes help control bowling-effects over large areas. Default: %(default)s')
 
-    parser.add_argument('--gps-z-offset',
-                        type=float,
-                        action=StoreValue,
-                        metavar='<float>',
-                        default=0,
-                        help='Set a GPS offset in meters for the vertical axis (Z) '
-                        'by adding it to the altitude value of the GPS EXIF data. This does not change the value of any GCPs. '
-                        'This can be useful for example when adjusting from ellipsoidal to orthometric height. '
-                        'Default: %(default)s')
-    
     parser.add_argument('--optimize-disk-space',
                 action=StoreTrue,
                 nargs=0,

opendm/dem/commands.py

@@ -1,5 +1,4 @@
 import os
-import subprocess
 import sys
 import rasterio
 import numpy
@@ -21,8 +20,6 @@ from opendm import log
 from .ground_rectification.rectify import run_rectification
 from . import pdal
 
-gdal_proximity = None
-
 try:
     # GDAL >= 3.3
     from osgeo_utils.gdal_proximity import main as gdal_proximity
@@ -30,13 +27,8 @@ except ModuleNotFoundError:
     # GDAL <= 3.2
     try:
         from osgeo.utils.gdal_proximity import main as gdal_proximity
-    except ModuleNotFoundError:
+    except:
-        # GDAL <= 3.0
+        pass
-        gdal_proximity_script = shutil.which("gdal_proximity.py")
-        if gdal_proximity_script is not None:
-            def gdal_proximity(args):
-                subprocess.run([gdal_proximity_script] + args[1:], check=True)
-
 
 def classify(lasFile, scalar, slope, threshold, window):
     start = datetime.now()

opendm/gpu.py

@@ -19,16 +19,13 @@ def has_popsift_and_can_handle_texsize(width, height):
             log.ODM_INFO("CUDA compute platform is not supported (detected: %s.%s but we need at least 3.5)" % (compute_major, compute_minor))
             return False
     except Exception as e:
-        log.ODM_WARNING(str(e))
+        log.ODM_INFO("Using CPU for feature extraction: %s" % str(e))
         return False
 
     try:
         from opensfm import pypopsift
-        if pypopsift.fits_texture(int(width * 1.02), int(height * 1.02)):
+        return pypopsift.fits_texture(int(width * 1.02), int(height * 1.02))
-            log.ODM_INFO("popsift can handle texture size %dx%d" % (width, height))
+    except (ModuleNotFoundError, ImportError):
-            return True
-        else:
-            log.ODM_INFO("popsift cannot handle texture size %dx%d" % (width, height))
         return False
     except Exception as e:
         log.ODM_WARNING(str(e))
@@ -84,12 +81,11 @@ def has_gpu(args):
             log.ODM_INFO("CUDA drivers detected")
             return True
         else:
-            log.ODM_INFO("No CUDA drivers detected")
+            log.ODM_INFO("No CUDA drivers detected, using CPU")
             return False
     else:
         if shutil.which('nvidia-smi') is not None:
             log.ODM_INFO("nvidia-smi detected")
             return True
         else:
-            log.ODM_INFO("No nvidia-smi detected")
             return False

opendm/gsd.py

@@ -5,7 +5,6 @@ import math
 from repoze.lru import lru_cache
 from opendm import log
 from opendm.shots import get_origin
-from scipy import spatial
 
 def rounded_gsd(reconstruction_json, default_value=None, ndigits=0, ignore_gsd=False):
     """
@@ -112,11 +111,15 @@ def opensfm_reconstruction_average_gsd(reconstruction_json, use_all_shots=False)
     with open(reconstruction_json) as f:
         data = json.load(f)
 
+    # Calculate median height from sparse reconstruction
     reconstruction = data[0]
-    points = np.array([reconstruction['points'][pointId]['coordinates'] for pointId in reconstruction['points']])
+    point_heights = []
-    tdpoints = points.copy()
+
-    tdpoints[:,2] = 0
+    for pointId in reconstruction['points']:
-    tree = spatial.cKDTree(tdpoints)
+        point = reconstruction['points'][pointId]
+        point_heights.append(point['coordinates'][2])
+
+    ground_height = np.median(point_heights)
 
     gsds = []
     for shotImage in reconstruction['shots']:
@@ -130,13 +133,6 @@ def opensfm_reconstruction_average_gsd(reconstruction_json, use_all_shots=False)
                 log.ODM_WARNING("Cannot parse focal values from %s. This is likely an unsupported camera model." % reconstruction_json)
                 return None
                 
-            shot_origin[2] = 0
-            distances, neighbors = tree.query(
-                shot_origin, k=9
-            )
-
-            if len(distances) > 0:
-                ground_height = np.median(points[neighbors][:,2])
             gsds.append(calculate_gsd_from_focal_ratio(focal_ratio, 
                                                         shot_height - ground_height, 
                                                         camera['width']))
@@ -149,6 +145,7 @@ def opensfm_reconstruction_average_gsd(reconstruction_json, use_all_shots=False)
     
     return None
 
+
 def calculate_gsd(sensor_width, flight_height, focal_length, image_width):
     """
     :param sensor_width in millimeters

opendm/orthophoto.py

@@ -13,7 +13,6 @@ from rasterio.mask import mask
 from opendm import io
 from opendm.tiles.tiler import generate_orthophoto_tiles
 from opendm.cogeo import convert_to_cogeo
-from opendm.utils import add_raster_meta_tags
 from osgeo import gdal
 from osgeo import ogr
 
@@ -167,7 +166,7 @@ def generate_tfw(orthophoto_file):
         log.ODM_WARNING("Cannot create .tfw for %s: %s" % (orthophoto_file, str(e)))
 
 
-def post_orthophoto_steps(args, bounds_file_path, orthophoto_file, orthophoto_tiles_dir, resolution, reconstruction, tree, embed_gcp_meta=False):
+def post_orthophoto_steps(args, bounds_file_path, orthophoto_file, orthophoto_tiles_dir, resolution):
     if args.crop > 0 or args.boundary:
         Cropper.crop(bounds_file_path, orthophoto_file, get_orthophoto_vars(args), keep_original=not args.optimize_disk_space, warp_options=['-dstalpha'])
 
@@ -180,8 +179,6 @@ def post_orthophoto_steps(args, bounds_file_path, orthophoto_file, orthophoto_ti
     if args.orthophoto_kmz:
         generate_kmz(orthophoto_file)
 
-    add_raster_meta_tags(orthophoto_file, reconstruction, tree, embed_gcp_meta=embed_gcp_meta)
-
     if args.tiles:
         generate_orthophoto_tiles(orthophoto_file, orthophoto_tiles_dir, args.max_concurrency, resolution)
 

opendm/osfm.py

@@ -296,8 +296,6 @@ class OSFMContext:
             config.append("matcher_type: %s" % osfm_matchers[matcher_type])
 
             # GPU acceleration?
-            if feature_type == "SIFT":
-                log.ODM_INFO("Checking for GPU as using SIFT for extracting features")
             if has_gpu(args) and max_dims is not None:
                 w, h = max_dims
                 if w > h:
@@ -307,12 +305,10 @@ class OSFMContext:
                     w = int((w / h) * feature_process_size)
                     h = int(feature_process_size)
 
-                    if has_popsift_and_can_handle_texsize(w, h):
+                if has_popsift_and_can_handle_texsize(w, h) and feature_type == "SIFT":
                     log.ODM_INFO("Using GPU for extracting SIFT features")
                     feature_type = "SIFT_GPU"
                     self.gpu_sift_feature_extraction = True
-                    else:
-                        log.ODM_INFO("Using CPU for extracting SIFT features as texture size is too large or GPU SIFT is not available")
             
             config.append("feature_type: %s" % feature_type)
 

opendm/photo.py

@@ -786,12 +786,6 @@ class ODM_Photo:
     def override_gps_dop(self, dop):
         self.gps_xy_stddev = self.gps_z_stddev = dop
 
-    def adjust_z_offset(self, z_offset):
-        if self.altitude is not None:
-            self.altitude += z_offset
-        else:
-            self.altitude = z_offset
-
     def override_camera_projection(self, camera_projection):
         if camera_projection in projections:
             self.camera_projection = camera_projection

opendm/rollingshutter.py

@@ -22,12 +22,6 @@ RS_DATABASE = {
     'hasselblad l2d-20c': 16.6, # DJI Mavic 3 (not enterprise version)
 
     'dji fc3582': lambda p: 26 if p.get_capture_megapixels() < 48 else 60, # DJI Mini 3 pro (at 48MP readout is 60ms, at 12MP it's 26ms) 
-    'dji fc8482': lambda p: (
-        16 if p.get_capture_megapixels() < 12 else  # 12MP 16:9 mode (actual 9.1MP)
-        21 if p.get_capture_megapixels() < 20 else  # 12MP 4:3 mode (actual 12.2MP)
-        43 if p.get_capture_megapixels() < 45 else  # 48MP 16:9 mode (actual 36.6MP)
-        58                                          # 48MP 4:3 mode (actual 48.8MP)
-    ), # DJI Mini 4 Pro (readout varies by resolution and aspect ratio, image heights all different)
 
     'dji fc350': 30, # Inspire 1
     

opendm/tiles/tiler.py

@@ -39,7 +39,7 @@ def generate_colored_hillshade(geotiff):
 
         system.run('gdaldem color-relief "%s" "%s" "%s" -alpha -co ALPHA=YES' % (geotiff, relief_file, colored_dem))
         system.run('gdaldem hillshade "%s" "%s" -z 1.0 -s 1.0 -az 315.0 -alt 45.0' % (geotiff, hillshade_dem))
-        system.run('"%s" "%s" "%s" "%s" "%s"' % (sys.executable, hsv_merge_script, colored_dem, hillshade_dem, colored_hillshade_dem))
+        system.run('%s "%s" "%s" "%s" "%s"' % (sys.executable, hsv_merge_script, colored_dem, hillshade_dem, colored_hillshade_dem))
         
         return outputs
     except Exception as e:

opendm/utils.py

@@ -1,12 +1,8 @@
 import os, shutil
 import numpy as np
 import json
-import rasterio
-from osgeo import gdal
-from datetime import datetime
-
 from opendm import log
-from opendm.photo import find_largest_photo_dims, find_mean_utc_time
+from opendm.photo import find_largest_photo_dims
 from osgeo import gdal
 from opendm.arghelpers import double_quote
 
@@ -118,41 +114,3 @@ def np_to_json(arr):
 
 def np_from_json(json_dump):
     return np.asarray(json.loads(json_dump))
-
-def add_raster_meta_tags(raster, reconstruction, tree, embed_gcp_meta=True):
-    try:
-        if os.path.isfile(raster):
-            mean_capture_time = find_mean_utc_time(reconstruction.photos)
-            mean_capture_dt = None
-            if mean_capture_time is not None:
-                mean_capture_dt = datetime.fromtimestamp(mean_capture_time).strftime('%Y:%m:%d %H:%M:%S') + '+00:00'
-
-            log.ODM_INFO("Adding TIFFTAGs to {}".format(raster))
-            with rasterio.open(raster, 'r+') as rst:
-                if mean_capture_dt is not None:
-                    rst.update_tags(TIFFTAG_DATETIME=mean_capture_dt)
-                rst.update_tags(TIFFTAG_SOFTWARE='ODM {}'.format(log.odm_version()))
-
-            if embed_gcp_meta:
-                # Embed GCP info in 2D results via
-                # XML metadata fields
-                gcp_gml_export_file = tree.path("odm_georeferencing", "ground_control_points.gml")
-
-                if reconstruction.has_gcp() and os.path.isfile(gcp_gml_export_file):
-                    gcp_xml = ""
-
-                    with open(gcp_gml_export_file) as f:
-                        gcp_xml = f.read()
-
-                    ds = gdal.Open(raster)
-                    if ds is not None:
-                        if ds.GetMetadata('xml:GROUND_CONTROL_POINTS') is None or self.rerun():
-                            ds.SetMetadata(gcp_xml, 'xml:GROUND_CONTROL_POINTS')
-                            ds = None
-                            log.ODM_INFO("Wrote xml:GROUND_CONTROL_POINTS metadata to %s" % raster)
-                        else:
-                            log.ODM_WARNING("Already embedded ground control point information")
-                    else:
-                        log.ODM_WARNING("Cannot open %s for writing, skipping GCP embedding" % raster)
-    except Exception as e:
-        log.ODM_WARNING("Cannot write raster meta tags to %s: %s" % (raster, str(e)))

requirements.txt

@@ -33,7 +33,7 @@ xmltodict==0.12.0
 fpdf2==2.4.6
 Shapely==1.7.1
 onnxruntime==1.12.1
-pygltflib==1.16.5
+pygltflib==1.15.3
 codem==0.24.0
 trimesh==3.17.1
 pandas==1.5.2

stages/dataset.py

@@ -234,9 +234,9 @@ class ODMLoadDatasetStage(types.ODM_Stage):
                                         item['p'].set_mask(os.path.basename(mask_file))
                                         log.ODM_INFO("Wrote %s" % os.path.basename(mask_file))
                                     else:
-                                        log.ODM_WARNING("Cannot generate mask for %s" % item['file'])
+                                        log.ODM_WARNING("Cannot generate mask for %s" % img)
                                 except Exception as e:
-                                    log.ODM_WARNING("Cannot generate mask for %s: %s" % (item['file'], str(e)))
+                                    log.ODM_WARNING("Cannot generate mask for %s: %s" % (img, str(e)))
 
                             parallel_map(parallel_sky_filter, sky_images, max_workers=args.max_concurrency)
 
@@ -338,10 +338,3 @@ class ODMLoadDatasetStage(types.ODM_Stage):
         if args.rolling_shutter and not reconstruction.is_georeferenced():
             log.ODM_WARNING("Reconstruction is not georeferenced, disabling rolling shutter correction")
             args.rolling_shutter = False
-        
-        # GPS Z offset
-        if 'gps_z_offset_is_set' in args:
-            log.ODM_INFO("Adjusting GPS Z offset by %s for all images" % args.gps_z_offset)
-
-            for p in photos:
-                p.adjust_z_offset(args.gps_z_offset)

stages/odm_dem.py

@@ -12,8 +12,6 @@ from opendm.cropper import Cropper
 from opendm import pseudogeo
 from opendm.tiles.tiler import generate_dem_tiles
 from opendm.cogeo import convert_to_cogeo
-from opendm.utils import add_raster_meta_tags
-
 
 class ODMDEMStage(types.ODM_Stage):
     def process(self, args, outputs):
@@ -89,8 +87,6 @@ class ODMDEMStage(types.ODM_Stage):
                     if pseudo_georeference:
                         pseudogeo.add_pseudo_georeferencing(dem_geotiff_path)
 
-                    add_raster_meta_tags(dem_geotiff_path, reconstruction, tree, embed_gcp_meta=not outputs['large'])
-
                     if args.tiles:
                         generate_dem_tiles(dem_geotiff_path, tree.path("%s_tiles" % product), args.max_concurrency, resolution)
                     

stages/odm_orthophoto.py

@@ -132,8 +132,7 @@ class ODMOrthoPhotoStage(types.ODM_Stage):
                     else:
                         log.ODM_INFO("Not a submodel run, skipping mask raster generation")
 
-                orthophoto.post_orthophoto_steps(args, bounds_file_path, tree.odm_orthophoto_tif, tree.orthophoto_tiles, resolution, 
+                orthophoto.post_orthophoto_steps(args, bounds_file_path, tree.odm_orthophoto_tif, tree.orthophoto_tiles, resolution)
-                    reconstruction, tree, not outputs["large"])
 
                 # Generate feathered orthophoto also
                 if args.orthophoto_cutline and submodel_run:

stages/odm_postprocess.py

@@ -1,8 +1,12 @@
 import os
+import rasterio
 
+from datetime import datetime
+from osgeo import gdal
 from opendm import io
 from opendm import log
 from opendm import types
+from opendm import photo
 from opendm.utils import copy_paths, get_processing_results_paths
 from opendm.ogctiles import build_3dtiles
 
@@ -13,6 +17,54 @@ class ODMPostProcess(types.ODM_Stage):
 
         log.ODM_INFO("Post Processing")
 
+        rasters = [tree.odm_orthophoto_tif,
+                    tree.path("odm_dem", "dsm.tif"),
+                    tree.path("odm_dem", "dtm.tif")]
+
+        mean_capture_time = photo.find_mean_utc_time(reconstruction.photos)
+        mean_capture_dt = None
+        if mean_capture_time is not None:
+            mean_capture_dt = datetime.fromtimestamp(mean_capture_time).strftime('%Y:%m:%d %H:%M:%S') + '+00:00'
+
+        # Add TIFF tags
+        for product in rasters:
+            if os.path.isfile(product):
+                log.ODM_INFO("Adding TIFFTAGs to {}".format(product))
+                with rasterio.open(product, 'r+') as rst:
+                    if mean_capture_dt is not None:
+                        rst.update_tags(TIFFTAG_DATETIME=mean_capture_dt)
+                    rst.update_tags(TIFFTAG_SOFTWARE='ODM {}'.format(log.odm_version()))
+
+        # GCP info
+        if not outputs['large']:
+            # TODO: support for split-merge?
+
+            # Embed GCP info in 2D results via
+            # XML metadata fields
+            gcp_gml_export_file = tree.path("odm_georeferencing", "ground_control_points.gml")
+
+            if reconstruction.has_gcp() and io.file_exists(gcp_gml_export_file):
+                skip_embed_gcp = False
+                gcp_xml = ""
+
+                with open(gcp_gml_export_file) as f:
+                    gcp_xml = f.read()
+
+                for product in rasters:
+                    if os.path.isfile(product):
+                        ds = gdal.Open(product)
+                        if ds is not None:
+                            if ds.GetMetadata('xml:GROUND_CONTROL_POINTS') is None or self.rerun():
+                                ds.SetMetadata(gcp_xml, 'xml:GROUND_CONTROL_POINTS')
+                                ds = None
+                                log.ODM_INFO("Wrote xml:GROUND_CONTROL_POINTS metadata to %s" % product)
+                            else:
+                                skip_embed_gcp = True
+                                log.ODM_WARNING("Already embedded ground control point information")
+                                break
+                        else:
+                            log.ODM_WARNING("Cannot open %s for writing, skipping GCP embedding" % product)
+
         if getattr(args, '3d_tiles'):
             build_3dtiles(args, tree, reconstruction, self.rerun())
 

stages/splitmerge.py

@@ -13,7 +13,7 @@ from opendm.cropper import Cropper
 from opendm.remote import LocalRemoteExecutor
 from opendm.shots import merge_geojson_shots, merge_cameras
 from opendm import point_cloud
-from opendm.utils import double_quote, add_raster_meta_tags
+from opendm.utils import double_quote
 from opendm.tiles.tiler import generate_dem_tiles
 from opendm.cogeo import convert_to_cogeo
 from opendm import multispectral
@@ -263,8 +263,7 @@ class ODMMergeStage(types.ODM_Stage):
 
                         orthophoto_vars = orthophoto.get_orthophoto_vars(args)
                         orthophoto.merge(all_orthos_and_ortho_cuts, tree.odm_orthophoto_tif, orthophoto_vars)
-                        orthophoto.post_orthophoto_steps(args, merged_bounds_file, tree.odm_orthophoto_tif, tree.orthophoto_tiles, args.orthophoto_resolution,
+                        orthophoto.post_orthophoto_steps(args, merged_bounds_file, tree.odm_orthophoto_tif, tree.orthophoto_tiles, args.orthophoto_resolution)
-                            reconstruction, tree, False)
                     elif len(all_orthos_and_ortho_cuts) == 1:
                         # Simply copy
                         log.ODM_WARNING("A single orthophoto/cutline pair was found between all submodels.")
@@ -306,8 +305,6 @@ class ODMMergeStage(types.ODM_Stage):
                         if args.tiles:
                             generate_dem_tiles(dem_file, tree.path("%s_tiles" % human_name.lower()), args.max_concurrency, args.dem_resolution)
                         
-                        add_raster_meta_tags(dem_file, reconstruction, tree, embed_gcp_meta=False)
-
                         if args.cog:
                             convert_to_cogeo(dem_file, max_workers=args.max_concurrency)
                     else:

winpostinstall.bat

@@ -1,18 +1,14 @@
 set ODMBASE=%~dp0
 set VIRTUAL_ENV=%ODMBASE%venv
-IF "%~1"=="" (set WRITABLE_VIRTUAL_ENV=%VIRTUAL_ENV%) ELSE (set WRITABLE_VIRTUAL_ENV=%~1)
+set PYENVCFG=%VIRTUAL_ENV%\pyvenv.cfg
-mkdir "%WRITABLE_VIRTUAL_ENV%"
+set SBBIN=%ODMBASE%SuperBuild\install\bin
 
 rem Hot-patching pyvenv.cfg
-set PYENVCFG=%WRITABLE_VIRTUAL_ENV%\pyvenv.cfg
+echo home = %ODMBASE%venv\Scripts> "%PYENVCFG%"
-echo home = %VIRTUAL_ENV%\Scripts> "%PYENVCFG%"
 echo include-system-site-packages = false>> "%PYENVCFG%"
 
 rem Hot-patching cv2 extension configs
-set SBBIN=%ODMBASE%SuperBuild\install\bin
+echo BINARIES_PATHS = [r"%SBBIN%"] + BINARIES_PATHS> venv\Lib\site-packages\cv2\config.py
-set CV2=%WRITABLE_VIRTUAL_ENV%\Lib\site-packages\cv2
+echo PYTHON_EXTENSIONS_PATHS = [r'''%VIRTUAL_ENV%\lib\site-packages\cv2\python-3.8'''] + PYTHON_EXTENSIONS_PATHS> venv\Lib\site-packages\cv2\config-3.8.py
-mkdir "%CV2%"
-echo BINARIES_PATHS = [r"%SBBIN%"] + BINARIES_PATHS> "%CV2%\config.py"
-echo PYTHON_EXTENSIONS_PATHS = [r'''%VIRTUAL_ENV%\lib\site-packages\cv2\python-3.8'''] + PYTHON_EXTENSIONS_PATHS> "%CV2%\config-3.8.py"
 
 cls