Image Classifier on 102 Oxford Flowers Dataset
Compile train.py for the training script and predict.py for the prediction script. Both are command line applications that run automatically upon compilation. Here are the following arguments each script takes:train.py
mode: either train or resume_train dir: directory of image files for training checkpoint: path to the checkpoint file model: default is vgg16 hidden_units: default is 4096 epochs: default is 30 lr: learning rate; default is 0.001 GPU: use GPU for training; default is false
predict.py
img_file: pass in image for inference checkpoint: checkpoint file mapping: category to name mapping file GPU: use GPU for prediction; default is false topk: predict top k classes; default is 1
See Jupyter Notebook for output.