A public database containing images and mask pair of X-ray CT slices
** UPDATE 17/4/2023 **
- TC001 mask and training data uploaded (.npy format)
- *D series is not available for download (TD,and MD)
Introduction
The dataset is stored in a Google Cloud Bucket. The Bucket contains files like MA001_train, MA001_mask, MA002_train, MA002_mask …. TC002_train, TC002_mask. Each file is a .npy file and can be loaded with the np.save function.
Each .npy file has either a “_train” or “_mask” suffix and the class it belongs to is written as it’s prefix. Files with “_train” suffix contains the training images in a (300, h, w) uint16 NumPy array. The bit-depth of each pixel is 16-bits (uint16).
Files with “_mask” contains the mask for the training image in a (300, h, w) uint8 NumPy array. The mask at index i represents the mask for ith training image in the training data. The mask is segmented in the following format: '''' ** 0 – Foreground 1 – Material 2 – Pore ** ''''
For Example, a file named “MA001_train” means it is the training image for class MA001. The corresponding mask can be found in the file named “MA001_mask”
Download Instructions
-
Go to Google Cloud Platform page: https://cloud.google.com/ . Click “Get started for free”
-
Follow the instructions on the screen until all three steps are completed.
-
Install Google Cloud SDK by clicking “Google Cloud CLI Installer” in blue from the following page: https://cloud.google.com/sdk/docs/install
-
After the .exe installer has finished downloading, execute the .exe installer and follow the instructions on the screen to complete the installation.
-
After the installation of the CLI is complete. Launch the CLI. A window similar to a command prompt window should appear.
-
Run the command “gcloud init”. When prompted for a numeric choice, input “1”. A network diagnostic should proceed automatically.
-
After the network diagnostic has completed successfully, input “Y” to login. A new window of your default internet browser should pop up, prompting a google login (similar to logging in to gmail).
-
After logging in at the pop-up browser, switch windows back to the cloud CLI. When prompted to pick a project, select the first project by inputting “1”. Take note that the project name will be different from the image below. Save the project name in its entirety somewhere safe.
-
The CLI is now setup
-
The dataset is roughly 6GB in size. To download the dataset, replace the parts in purple and run the command to download the entire dataset:
gsutil -u PROJECT-NAME cp -r gs://datatest_13/XCTPore SAVE_DIRECTORY
Warning: The command above will charge about USD$0.50 to your GCP account if it is executed without error. Any downloads will incur charges to your GCP account. It is to do the following before proceeding: • Ensure a stable internet connection • Download the entire dataset and keep a backup of the original dataset on your local machine to minimize charges.
CONTRIBUTING TO THE DATASET To contribute to dataset, run the following command:
gsutil -u PROJECT-NAME cp -r DATASET_DIRECTORY gs://datatest_13/upload_xctpore