nbutter/megadetector repository overview

⁠MegaDetector 5 Container

Docker/Singularity image to run Megadetector 5⁠ on Centos 6.9 kernel (Ubuntu 16.04) with GPU support.

Cuda, torch, etc versions have been modified from the original conda environment⁠. Other combinations may work.

Check the detail on the github repo⁠

If you have used this work for a publication, you must acknowledge SIH, e.g: "The authors acknowledge the technical assistance provided by the Sydney Informatics Hub, a Core Research Facility of the University of Sydney."

⁠Build with docker

sudo docker build . -t nbutter/megadetector:ubuntu1604

⁠Run with docker.

To run this, mounting your current host directory in the container directory, at /project, and execute a run on the test images (that live in the container) run:

sudo docker run --gpus all -it -v `pwd`:/project nbutter/megadetector:ubuntu1604 /bin/bash -c "cd /project && python /build/cameratraps/detection/run_detector_batch.py /build/blobs/md_v5a.0.0.pt /build/cameratraps/test_images/test_images/ mdv4test.json --output_relative_filenames --recursive"

⁠Push to docker hub

sudo docker push nbutter/megadetector:ubuntu1604

See the repo at https://hub.docker.com/r/nbutter/megadetector⁠

⁠Build with singularity

sudo singularity build mega.img docker://nbutter/megadetector:ubuntu1604

⁠Run with singularity

To run the singularity image (noting singularity mounts the current folder by default)

singularity run --nv mega.img /bin/bash -c "python /build/cameratraps/detection/run_detector_batch.py /build/blobs/md_v5a.0.0.pt /build/cameratraps/test_images/test_images/ mdv4test.json --output_relative_filenames --recursive"