An ensemble model created from ResNet, VGGNet, DenseNet, Inception, EfficientNet and MobileNet instances weighted with a logistic regressor, to classify images of currencies of 211-different classes. Winning entry for the https://www.kaggle.com/competitions/currency-prediction-challenge with around 88% accuracy.
To run this on the dataset found at https://www.kaggle.com/competitions/currency-prediction-challenge/data first train the component models, and load the weights of the 6 models.