1. Data acquisition and processing
Using the CIFAR-10 [2] dataset, which is provided under the MIT license.
The dataset consists of 60,000 32x32 pixel RGB images divided into 10 different categories. It is split into 50000 training samples and 10000 testing samples and is perfectly balanced, which means the dataset contains 6000 images for each class.
Datasets can be easily loaded by doing the following:
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()The dataset contains the following classes: airplane, car, bird, cat, deer, dog, frog, horse, boat, and truck. Two tasks for the multi-task model to learn will be classification on these labels, see:
Task 1: Multi-category classification on the modified CIFAR10 dataset (airplane, car, bird, cat, dog, frog, boat, and truck labels, modification instructions below).
Task 2: Binary classification (labels for animals and vehicles).