Tensorflow Binary Classification using Sigmoid. Oversampling and Undersampling to balance dataset and improve accuracy.
Oversampling: Increase the number of instances in the minority class (sneaker class) by duplicating or generating synthetic samples. Undersampling: Decrease the number of instances in the majority class (non-sneaker class) by randomly removing samples until both classes have approximately the same number of instances.