Flower Classifier

This is a flower classifier for the kaggle competition. The best accuracy is currently 85%.

How to use the model

Install required packages

boto, boto3, botocore, numpy, pillow, tensorflow (1.4.0)

sh requirements.sh

Download dataset and trained model

python download.py

1. Download flower images from the link: https://www.kaggle.com/alxmamaev/flowers-recognition
2. Download vgg16_weights.npz from https://www.cs.toronto.edu/~frossard/post/vgg16/
3. Move it under the depository folder
4. Resize and split the image data into training and testing set. Create the corresponding TFRecord to reduce RAM requirement

python build_TFRecord.py    --mode [train, test] \
                            --partition_size [1000]

Train the model

python main.py --train True \
               --test False \
               --epoch number_epoch_to_be_trained \
               --learning_rate learning_rate_in_training \
               --checkpoint_dir directory_to_save_model 

Test the model

python main.py --train False \
               --test True \
               --epoch number_epoch_to_be_trained \
               --learning_rate learning_rate_in_training \
               --checkpoint_dir model_directory

Use the model to classify images

python main.py --input_images_dir input_images'_directory \ 
               --checkpoint_dir model_directory 

--input_images_dir flag can receive a directory as well as the path to a single image.

Have fun with flowers!