Skip to content

Latest commit

 

History

History
122 lines (82 loc) · 4.09 KB

readme.md

File metadata and controls

122 lines (82 loc) · 4.09 KB
Raw

CARES Dockers

Docker images supporting the Nvidia CUDA toolkit have been created for internal purposes.

Image-Segmentation

Read the instruction here.

ROS Noetic + GPU

Read the instruction here.

ROS Noetic + Detectron2

Read the instruction here.

Kaggle Docker Python

Read the instruction here.

ROS2 Humble + Gazebosim

WIP

Ubuntu20.04 + Cuda11.6

WIP

(Deprecated) Pytorch 1.13 (Cuda 11.6)

The image was built based on the Nvidia image and installed PyTorch 1.13 + CUDA 11.6 in a Conda environment. To use it, you will need to first pull it from Docker Hub.

# You should execute the following command on the workstation.
docker pull mycares/pytorch:1.13.0-cuda11.6-ubuntu20.04

Next, start a container by running the following command.

# You should execute the following command on the workstation.
docker run -it --runtime=nvidia -v /dev/shm:/dev/shm --mount source=datastore,target=/data mycares/pytorch:1.13.0-cuda11.6-ubuntu20.04 /bin/bash

After connecting to the container, activate the Conda environment for PyTorch.

# You should execute the following command within the container.
conda activate torch1.13.0

It is recommended to use pip to install Python packages.

(Deprecated) Detectron2 (Cuda 11.6)

The image was built based on the Nvidia image and installed PyTorch 1.13 + CUDA 11.6 + Detectron2(2023.03.21) in a Conda environment. To use it, you will need to first pull it from Docker Hub.

# You should execute the following command on the workstation.
docker pull mycares/detectron2:2023.03.21

Next, start a container by running the following command.

# You should execute the following command on the workstation.
docker run -it --runtime=nvidia -v /dev/shm:/dev/shm --mount source=datastore,target=/data mycares/detectron2:2023.03.21 /bin/bash

Once connected to the container, activate the conda environment for detectron2.

# You should execute the following command within the container.
conda activate torch1.13.0

You can try a quick demo of Detectron2 by copying any image file (and naming it 'input.jpg') into the '/home/$USER/data' directory of the workstation using FileZilla Client, and then listing the file in your container. In this example, we will use the following image.

input image

List the file in your container.

# You should execute the following command within the container.
ls /data

In order to perform a quick demo, navigate to the '~/workspace/detectron2/demo' directory after verifying that the input.jpg file has been successfully transferred to your container.

# You should execute the following command within the container.
cd ~/workspace/detectron2/demo

Create a folder named 'output' to save the output files from the demo.

# You should execute the following command within the container.
mkdir ~/workspace/detectron2/demo/output

Now, we are ready to run the demo script.

# You should execute the following command within the container.
python demo.py --config-file ../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml \
  --input /data/input.jpg \
  --output output \
  --opts MODEL.WEIGHTS detectron2://COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl

Copy the output folder into /data in your container to see the result.

# You should execute the following command within the container.
cp -r output/ /data/

Once you have refreshed FileZilla Client, go to the shared directory (/home/$USER/data) on the workstation. Inside, you'll find a directory named 'output' that contains the resulting file.

output image

Please utilize pip within this conda environment if you require additional Python packages.