python - How do you see how accurate TensorFlow an image classification model is for each class?

Question

I'm following through the image classification tutorial on the tensor flow website: https://www.tensorflow.org/tutorials/images/classification

The model classifies flowers into one of 5 classes: daisy, dandelion, roses, sunflower and tulips.

I can see what the overall accuracy is, but is there any way I can know how accurate it is for each class?

For example, my model could be very good at predicting daisies, dandelions, roses, and sunflowers (near 100% accuracy), and poor at tulips (near 0%) and I think I'd still see 80% overall accuracy (assuming the classes are balanced). I'd need to know the accuracy for the individual classes to differentiate that performance from a model that predicts all classes at an approximately equal 80% accuracy.

Answer 1

You could do that simply by using classification report in sklearn.

Refer Documentation