2023 Grand Challenge on Objective Quality Metrics for Volumetric Contents

This repository contains the files for building and running the container for the Objective Quality Metrics Grand Challenge.

Building the container

The container can be built as such:

docker build -t grandchallenge .

After that, the built container is available under the tag grandchallenge:latest. The container can also be built and tagged by calling the script build.sh:

bash build.sh

Building the base image

Keep in mind that the image is based on the mathworks/matlab:r2022a image with the Computer Vision and the Statistics and Machine Learning toolboxes installed. To create this base image, run the following command:

docker run --init -it -p 5902:5901 -p 6081:6080 --shm-size=512M mathworks/matlab:r2022a -vnc

Then, in a web browser, go to http://localhost:6081 and install the required toolboxes in a Matlab instance launched using sudo from a terminal. With the container still running, tag a new image:

docker commit --change 'ENTRYPOINT ["/bin/run.sh"]' [IMAGE_ID] grandchallenge_toolboxes:r2022a

Where [IMAGE_ID] is the container ID of the running Matlab container. After that be base image can be built using the command at the beginning of this section.

Running the container

To run the container and obtain the results, run the following command:

docker run -v [DATASET_DIR]:/app/dataset/ -v [MODEL_DIR]:/app/model/ grandchallenge python test.py

Where [DATASET_DIR] is the path to the directory containing the data set and [MODEL_DIR] is the path to the directory containing the model. DATASET_DIR should contain two subdirectories named dis/ and ref/, containing the distorted and reference point clouds respectively.

After the container has finished running, the results can be extracted using the following command:

docker cp [IMAGE_ID]:/app/results.csv results.csv

Where [IMAGE_ID] is the ID of the container that computed the results.

For convenience, there is also the script run.sh, which performs all these steps in sequence and performs cleanup afterwards. The script can be invoked as follows, passing in the absolute paths of the dataset directory and the model directory:

bash run.sh dataset_dir model_dir

After a completed run, the predictions are available in the current working directory as results.csv.