I have tried to add a class to the dogs versus cats notebook of lesson 1. It seems to work, but several graphs are blank and I don’t understand why.
I randomly picked class which I thought would be very different from the two existing animal classes - “road signs”. I figured that signs would be flat, often boxy, mostly unicolor, have text and the whole not alive part. The results are not as good as I would hope, but that might be because I used images from google images and the minimal sample (the one names “sample”) subset of images the cats-versus-dogs dataset.
Here is a PDF xelatex output of the notebook (I assume this is the most convenient):
lesson1-DogsCatsSigns-Copy1.pdf (2.7 MB)
Here is the notebook:
scratch that, it isn’t a format that is allowed.
Specific Questions:
- Why is
learn.sched.plot_lr()
graph blank (page 12 of PDF print out)? - Why is
learn.sched.plot()
graph blank (also page 12)? - Why does
learn.sched.plot_lr()
on page 14 look like a rectifier? - What does
<matplotlib.figure.Figure at 0x7f0b1c639dd8>
mean and why can’t I graphplot_val_with_title(most_by_correct(0, False), "Most incorrect cats")
or the dog one on page 18???
Any help would be very much appreciated. Thanks!