This repository contains 2 models:
Our 3rd and 4th research models mmsegmentation
and DDRNet, storing the programs for
each model and using Docker containers to create a consistent environment for them to be run in. Documentation for the individual
models are stored in their respective folders.
Individual information provided in the model sections. Please install Docker through normal linux pathways.
sudo apt-get update
sudo apt-get install apt-transport-https ca-certificates curl gnupg lsb-release
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
sudo apt-get update
sudo apt-get install docker-ce docker-ce-cli containerd.io
sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker
1.Error massages: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
run this command to give the AUTHENTICATION.
systemctl start docker
- โSub-process /usr/bin/dpkg returned an error code (1)โ error when Install Docker Engine, containerd, and Docker Compose.
run this command to Reconfiguring the database.
sudo dpkg --configure -a
- Dockerfile: this file will set up a docker image
- Create a image to run a Dockerfile in directory: docker/
docker build -t {IMAGE_NAME} docker/
For instance:
docker build -t mmsegmentation docker/
- List images to check if the image has been created:
docker images
- Host data: the data in the host that export the data into docker image
- Docker: data: import the data from the host as a data directory
docker run --gpus all --shm-size=8g -it -v {Host: data}:{Docker: data} {IMAGE_NAME}
For instance:
docker run --gpus all --shm-size=8g -it -v /home/usyd-05a/data:/mmsegmentation/data mmsegmentation
! There are few basic packages in Docker, not all of them have been installed. This means you may need to install some packages in the container.
-
Models are first trained, then evaluated. All models will use mean IOU (mIOU) as a measure of their accuracy, defined by the overlap between predicted out and target segmentation mask.
-
The dataset is downloaded from official website
+ NOTE: Any errors with memory, for example pertaining to CUDA, is foremostly likely due to insufficient computing resources on the server. Please try again when there is little usage to see if the issue is resolved.
- We use Tensorboards to visulize scalars (follow instructions here to install tensorboards
As the models are predeveloped and obtained online, code style is not applicable. However any Python files created by the group will use PEP 8.
README presented in the Benchmarking folder