This repository hosts a Colab notebook which trains a model to recognise flowers from Oxford's 102 Cateogry Flower Dataset
I use a convolutional neural network with cyclical learning rates in order to achieve an accuracy of 97.3% on the validation set. This far exceeds the 72.8% accuracy achieved with featurization approaches from the original paper for this dataset.