search

fastai / course-v3

The 3rd edition of course.fast.ai
https://course.fast.ai/
Apache License 2.0
4.91k stars 3.57k forks source link

Getting wrong tags #514

Closed guixermo closed 4 years ago

guixermo commented 4 years ago

I downloaded a dataset consisting on fruits. From that dataset i only took 4 classes to diferentiate. The problem comes when i plot the results of the classification with plot_multi_tp_losses(...) The predicted label is correct but the actual label is wrong or some of them are empty. See the imaged for an example. image image image

the previous classification by folders seems to be right data.show_batch image

guixermo commented 4 years ago

I was using interp.plot_multi_top_losses which i don't know why, it shows wrong labels. Using plot_top_losses show correct values.