My only requirement is to use odm for orthophoto generation. With a cpu it takes too long , more than 30mins. I tried using the --gpus all flag , but it detects the nvidia gpu but it does not use it. Any idea on how to use gpu for faster gneration of orthophotos.
Can you give some details on how you ran and share your log?
@smathermather
Log(Running with GPU): https://drive.google.com/file/d/1KDPRvRmH_ZReZHvfOmOp6w7nPXaDBB-y/view?usp=sharing
COmmand I ran
time docker run --rm --gpus all -v /home/felix/odm_sance:/datasets opendronemap/odm:gpu --project-path /datasets batch1 --fast-orthophoto --skip-report --skip-3dmodel --feature-type sift --feature-quality high --max-concurrency 4 --orthophoto-resolution 5 2>&1 | tee /home/felix/odm_sance/batch1/run_gpu.log
This part of logs might be more helpful
e[39m[INFO] no_gpu: Falsee[0m
e[39m[INFO] Checking for GPU as using SIFT for extracting featurese[0m
e[39m[INFO] nvidia-smi detectede[0m
e[93m[WARNING] cannot import name ‘pypopsift’ from ‘opensfm’ (/code/SuperBuild/install/bin/opensfm/opensfm/init.py)e[0m
e[39m[INFO] Using CPU for extracting SIFT features as texture size is too large or GPU SIFT is not availablee[0m
I have a nvidia 3050ti with 4gb vram and I have 156 images in total, i try to run odm on a batch of 52 images at a time
I was able to solve this using the tag odm:gpucore but it is too old, the last update was 4 years ago, also the odm supports gpu acceleration for sift only, is it true? My only goal is to generate an orthophoto , so how do we speed up, I have used fast-orthophoto