search

raghakot / keras-vis

Neural network visualization toolkit for keras
https://raghakot.github.io/keras-vis
MIT License
2.98k stars 660 forks source link

Where to find implementation of `visualize_class_saliency` and `visualize_regression_saliency`? #169

Open frederikfaye opened 5 years ago

frederikfaye commented 5 years ago

From https://github.com/raghakot/keras-vis/issues/38#issuecomment-311859428:

visualize_xxx now has a regression and class variant separated out.

However, nothing like this is in vis/visualization/activation_maximization.py, and in https://github.com/raghakot/keras-vis/blob/668b0e11dab93f3487f23c17e07f40554a8939e9/vis/visualization/saliency.py#L54-L55 it is only referred to, but not implemented anywhere.

Where can I read about visualize_class_saliency and visualize_regression_saliency?

PoornimaRamaraj commented 5 years ago

I'm looking for visualize_regression_saliency.

keisen commented 5 years ago

Hi there.

visualize_class_saliency and visualize_regression_saliency don't exist. (I think, at first of developing keras-vis, there was possibility that they was created. I don't know the details, but, at least, they have not existed now.)

If you want to visualize saliency for outputs of regression model, you could use visualize_saliency or visualize_saliency_with_losses. And then the following links should help you.

https://raghakot.github.io/keras-vis/visualizations/saliency/#regression-dense-layer-visualization

and

https://github.com/raghakot/keras-vis/blob/668b0e11dab93f3487f23c17e07f40554a8939e9/applications/self_driving/visualize_attention.ipynb