Plant seedling classification is a competition on Kaggle, which requires to classify unseen seedlings' pictures into one of the twelve given species classes accurately. We intended to apply convolutional neural networks (CNN) and residual networks to perform the task.
Coding environment : Google Colab with GPU supported
Package : pandas, numpy, matplotlib.pyplot (graphing and ploting result), os, sys, cv2, keras.model, keras.layer, keras.callback, keras.preprocessing.image (for image data generator), keras.utils, sklearn.metrics, sklearn.model_selection (for train_test_split), tqdm, glob, tensorflow, torch, torchvision, glob, PIL, seaborn
hyperparameters : scale = 100, seed = 7
Both of us contribute a lot to this project. Pei-fan Liu did the part of CNN and Alexnet, and Pei-Chen Ho was responsible for the implementation of the transfer models and SVM. We work together on the report and the literature researches.