Visualization Task 3

Because this specific run only ran through 1 epoch and achieved an accuracy of only 88%, we decided it would make for the best portrail for our histogram.
The table was grouped by "prediction" column and as you can see when clicking the top left "pyramid" button, this table show's all the images that predicted 9's wrongly.
As you can see by the column "image.count" the model predicted it wrongly 31 times and on the histogram you can what the actual number was and how many times a picture of a specific number was predicted wrongly, 8 being the highest with 16 wrong predictions.

For the confusion matrix we ran over multiple tests to show where our model had False Predictions (FP, FN) and it gives a bit of an overview on what numbers our model seemed to have difficulties in distinction.