• In this project I train an image classifier to recognize different species of flowers among 102 flower categories.
• Use the trained classifier to predict image content of top-K predictions
• Develope a Command Line Application to allows users to set hyperparameters for learning rate, number of hidden units, training epochs, GPU/CPU

• Python 3.x
• Numpy
• PyTorch
• PIL
• json
• matplotlib
• collections

• Jupyter notebook file: Image Classifier Project_july01_2018_submit
• html file of Jupyter notebook: Image Classifier Project_july01_2018_submit.html
• Image folder (assets): containing figures in the notebook
• A json file: cat_to_name.json containing a dictionary that translate folder name to flower name
• A helper file: workspace_utils.py containing active_session function that used to avoid disconnecting program during long run
• Two Python file: train_submission.py and predict_submission.py for train and prediction using command line for Image classifier
• Two txt file : sample_train_in_out.txt and sample_predict_in_out.txt containg a successful printout of train.py and predict.py in the output
• MIT License file

• Clone the repository use: git clone https://github.com/ania4data/Developing_AI_application_for_classification.git

Train the network on the dataset

• Basic usage: python train_submission.py data_directory
• Prints out training loss, validation loss, and validation accuracy as the network trains
• Set directory to save checkpoints: python train_submission.py data_dir --save_dir save_directory
• Choose architecture: python train_submission.py data_dir --arch "resnet152"
• Set hyperparameters: python train_submission.py data_dir --learning_rate 0.01 --hidden_units 512 --epochs 20
• Use GPU for training: python train_submission.py data_dir --gpu

Predict flower name from an image with predict_submission.py along with the probability of that name.

• Basic usage: python predict_submission.py /path/to/image checkpoint_from_training
• Return top K most likely classes: python predict_submission.py input checkpoint_from_training --top_k 3
• Use a mapping of categories to real names: python predict_submission.py input checkpoint_from_training --category_names cat_to_name.json
• Use GPU for inference: python predict_submission.py input checkpoint_from_training --gpu

The zipped 102flowers.tgz file can be downloaded from download section of Visual Geometry Group in Oxford University website:

• Link to the study