This is a course project for Introduction to Data Science, School of Mathematical Sciences, Peking University, 2018 autumn.
In this flower classification task, we need to train our model based on a total of 3,899 images of five species of flowers, and give the type of 424 flowers in the test set to get as much accuracy as possible.
Kaggle competition link: https://www.kaggle.com/c/datascience2018
Our team (12138) finally achieve 0.97641 (3rd) in publicboard and 0.95754 (3rd) in privateboard
Results and more details in report.
DOWNLOAD TRAIN & TEST SET FROM THE COMPETITION LINK AND PRE-TRAINED WEIGHTS FOR MODELS BEFORE RUNNING!