CZ4041-Machine-Learning

This project aims to complete the Leaf Classification challenge in Kaggle.

Dependencies

• Tensorflow
• Keras
• OpenCV

Folders

• models: Contains code for different models experimented.
• tools: Contains code for feature extraction etc.

Feature Extraction

Besides the features provided by Kaggle dataset. Other features were extracted for experiments.

1. In cv_feature_extraction.py, OpenCV was used to extract extra features. Each feature was implemented in one function.
2. In time-series_feature_extraction.ipynb, a new kind of feature was experimented but did not produce good result in further training.

Models

Several model architectures were proposed and experimented.

1. In baseline.py, only features provided by Kaggle dataset were used. It is trained as a baseline result for further experiments.
2. In train_manual_extracted_feature.py, features extracted by OpenCV (cv_featue_extraction.py) were added for regression.
3. In densenet121.py, the DenseNet architecture was applied for training the dataset.
4. In capsule_net.py, the CapsuleNet architecture was applied for training the dataset.
5. In just_conv.py, a model with simple convolutional layers followed by fully connected layers were trained and evaluated.
6. In autoencoder folder, a model with a convolutional autoencoder followed by fully connected layers were trained and evaluated. ae.py file implemented the convolutional autoencoder and model.py file implemented the whole model including dense layers.