This project is only for educational purposes. I did it while I was following the Udacity DataScience nanodegree.
It is a supervised learning classification problem in which you have to develop code for an image classifier built with
PyTorch. It is firstly done with a notebook and then we convert it into a command line application.
The goal is to build a new classifier by using already existing models and replacing the last classifier layer by a new
one we train for a specific task (transfer learning).
Goal of this project is to classify flowers images among 102 possible classes.
Please refer to the notebook for more details and step by step discovery.
This project requires Python 3.6 and the following Python libraries installed:
- NumPy (I used 1.16.4)
- Pandas (I used 0.24.2)
- matplotlib (I used 3.1.0)
- pytorch (I used 1.1.0)
- pillow (PIL fork, I used 6.0.0)
- iPython Notebook to run the notebook
There are 2 scripts: train and predict, obviously the first one to train a model and the second one to load a previously trained model and classify an image.
Train a new network on a data set with train.py
Basic usage: python train.py data_directory
Prints out training loss, validation loss, and validation accuracy as the network trains
Options:
- Set directory to save checkpoints:
python train.py data_dir --save_dir save_directory - Choose architecture:
python train.py data_dir --arch "vgg13" - Set hyperparameters:
python train.py data_dir --learning_rate 0.01 --hidden_units 512 --epochs 20 - Use GPU for training:
python train.py data_dir --gpu
Sample:
python train.py ./flowers --arch vgg16 --gpu --save_dir ./ouput
Predict flower name from an image with predict.py along with the probability of that name.
That is, you'll pass in a single image /path/to/image and return the flower name and class probability.
Basic usage: python predict.py /path/to/image checkpoint
Options:
- Return top KK most likely classes:
python predict.py input checkpoint --top_k 3 - Use a mapping of categories to real names:
python predict.py input checkpoint --category_names cat_to_name.json - Use GPU for inference:
python predict.py input checkpoint --gpu
Sample:
python predict.py ./flowers/valid/28/image_05265.jpg output/checkpoint-script.pth --top_k 5 --category_names cat_to_name.json --gpu
It then generates a picture like this one:
