You will build an image classifier from scratch that will identify different species of flowers.
The data set contains around 7000 images of flowers from 102 different species. The provided images were split into a training set and a validation set. You can download the images from here as a zipped archive. Just uncompress it and you should be good to go.
It was possible to obtain an accuracy of 98.9% using a pretrained Resnet 152 model on the provided validation set after 27 epochs (18 epochs with only the classifier parameters unfrozen and 9 with all model parameters unfrozen).
This project requires Python 3.7 and the following Python libraries installed:
You will also need to have software installed to run and execute a Jupyter Notebook
If you do not have Python installed yet, it is highly recommended that you install the Anaconda distribution of Python, which already has some of the above packages and more included.
Code is provided in the Image Classifier Project_Resnet152_newflowers.ipynb notebook file. If you are interested in how the parameters influence the outcome, please feel free to explore this Python file.
In a terminal or command window, navigate to the top-level project directory pytorch_challenge/ (that contains this README) and run one of the following commands:
ipython notebook Project_Resnet152_newflowers.ipynbor
jupyter notebook Project_Resnet152_newflowers.ipynbThis will open the Jupyter Notebook software and project file in your browser.