Getting more details about image classification models
In the Training a classification model with a simple curated vision dataset recipe of Chapter 6, Training Models with Visual Data, you trained an image classification model using the CIFAR curated dataset. The code to train and exercise the model was straightforward because we took advantage of the highest-level structures in fastai. In this recipe, we will revisit this image classification model and explore techniques in fastai to get additional information about the model and its performance, including the following:
- Examining the pipeline that fastai generates to prepare the data
- Getting a chart of the training and validation loss during the training process
- Displaying the images where the model performs worst
- Displaying the confusion matrix to get a snapshot of where the model is not doing well
- Applying the model to the test set and examining the model's performance on the test set
In this...
