I’m using PyODM to generate orthomosaic images from a large set of TIFF files (ranging from 1,000 to 5,000 images) captured by a drone. However, I notice that the generated orthomosaic image is different each time I run the process, even though I’m using the same set of images and the same code.
Setup:
- Repository: I’m using the OpenDroneMap repository cloned as follows:
# Clone the OpenDroneMap repository git clone https://github.com/OpenDroneMap/ODM --recursive cd ODM - Code: Here’s a simplified version of my Python code using PyODM:
import threading
from pyodm import Node, exceptions
import glob
import os
from pathlib import Path
def progress_callback(task_id):
if int(task_id == 100):
print("Uploading Images done")
def print_progress(task):
taskInfo = task.info()
print(taskInfo.status)
def startOrthoProcess():
node = Node('localhost', port=3000)
try:
print(node.info())
except exceptions.NodeConnectionError as e:
print("Cannot connect: " + str(e))
return {
"status": 500,
"message": "Cannot connect to Node"
}
option_flags = {
'auto-boundary': True,
'radiometric-calibration': "camera"
}
image_files = [str(p) for p in Path("/path/to/images").rglob("*.tif")]
print(len(image_files))
task = node.create_task(image_files, option_flags, progress_callback=progress_callback)
taskInfo = task.info()
print(taskInfo)
timer = threading.Timer(60.0, print_progress, args=[task]) # 60 second timer to show progress
timer.start()
try:
result_dir = "results"
if not os.path.exists(result_dir):
os.makedirs(result_dir)
task.wait_for_completion()
task.download_assets(result_dir)
print(f"Result saved in: {result_dir}/odm_orthophoto/odm_orthophoto.tif")
return {
"status": 200,
"message": "Successfully generated Orthomosaic"
}
except exceptions.TaskFailedError as e:
print("Error: " + str(e))
return {
"status": 500,
"message": str(e)
}
def main():
startOrthoProcess()
if __name__ == "__main__":
main()
What I’ve Tried:
- I’ve ensured that the images and metadata are consistent.
- The
option_flagsare the same for each run. - I’ve checked the server environment, and it seems stable.
- The variability persists even when processing 1,000 to 5,000 images.
Questions:
- What could cause the orthomosaic images to differ each time the process is run?
- Are there any specific options or flags in OpenDroneMap that might introduce variability?
- How can I ensure that the results are consistent across multiple runs?
Any insights would be greatly appreciated!