Datasets

At this stage, AI4HPC supports these 2 exemplary datasets.

1. Actuated TBL

2. MNIST

Actuated TBL dataset is used for major development, whereas MNIST is for minor tests.

Actuated TBL dataset

The Actuated Turbulent Boundary Layer (ATBL) dataset is generated in the compute-driven use-case Task 3.1 “AI for turbulent boundary layers”. Briefly, the training dataset is generated using a high-resolution Large-Eddy Simulation (LES) approach to simulate the flow over a moving geometry. This creates oscillating boundary layers which reduces the friction drag of the turbulent boundary layer.

Download links with further information is here.

MNIST dataset

The Modified National Institute of Standards and Technology MNIST consists of a collection of handwritten digital images used for character recognition. This dataset is an extension of the original dataset available from the National Institute of Standards and Technology (NIST). The dataset consists of 60,000 sample digital images for training and 10,000 samples for testing purposes. Each sample image is centered and represented in fixed 28x28 black and white pixels. The grayscale levels are introduced to each image using anti-aliasing techniques.