search

facebookresearch / odin

A simple and effective method for detecting out-of-distribution images in neural networks.
Other
526 stars 102 forks source link

About the parameters δ, T, ε #18

Open muskahya opened 3 years ago

muskahya commented 3 years ago

Hi, The parameters are tuned by using ood val data. Can we generalize the tuned parameters for the other ood data? Or when we have new ood samples do we have to tune the parameters again?

Also, I am curious about the possible usage of this technique on signal data like radar signals.

Thanks

YixuanLi commented 3 years ago

Hi @muskahya, yes we have done the ablation in the paper, where you can see the performance on other OOD datasets: https://arxiv.org/abs/1706.02690.

Alternatively, you might want to use our latest energy-based approach, which is hyperparameter free and gives almost the same performance as ODIN: https://arxiv.org/abs/2010.03759

Hope it helps!