Will "Multi Label Classification Interpretation" be in v2?

fastai fastai dev
1

Hi everyone !

I’m creating a series of notebooks for the students in the fastai study group in Lille (northern France). So far they are based on part1 v1 from last year (as I understand it, the high level api shouldn’t change too much).

While creating exercises for multilabel classification, I discovered that MultiLabelClassificationInterpretation is not implemented !

So I’m thinking about making this my first contribution to the library, but before I dive into the specifics, I’d like to know whether this is currently being developped for v2. Is anybody working on this, or will my contrib be useful if I take the time to do it ?

Have a nice day :wink:

1 Like
2

If you look at the notebook 23, there is already something for multi label in terms of interpretation. It’s all defined in the dispatched plot_top_losses methods inside vision.learner.

2 Likes
3

I’ll take a look, thank you very much !

4

Thank you @sgugger ! I tested this on the planet dataset and works like a charm. :smiley:

PS : Your blog is awesome!

1 Like