Hey guys
I’ve been noticing a trend in my 3 days of Fastai: almost every time I take part in making a dataset I seem to fall into divergence (not overfitting I think since that should be a rare thing) and more often than not, bad error_rates (got my first good one in my 5 try at the previous assignment, so I guess that’s natural).
Per example: in this lesson two there’s the example of the bears, so I decided to do the same thing, only the pictures where different (unless there was a major coincidence) besides that divergence when doing the download o csv file, everything I did mirrored the notebook/class. Yet:
For comparison, the class one was:
My dataset seems to be half the size of the one used in class:
And then when plotting I’m already used to see things like this:
When the class one was like:
This seems horrendously far from what they got Did anyone get similar results? Or is this probably something I’m doing wrong?