Is there any post-hoc easily applicable explainable AI tool for detections carried out by complex CNN-base architectures e.g. Mask R-CNN?
yes Marilyn Bello you can search here: https://euraxess.ec.europa.eu/
Excuse me, Qamar Ul Islam but what does that have to do with my question?
As such, there is no explainable AI(XAI) tool or package is readily available mainly because of the following reasons,
It is mostly application-specific and secondly, this domain has not explore much due to the fact that we are more focused on what results we are achieving through CNN or another Deep Neural model rather than XAI.
But, still, there are many research journals available in this regard.
The following Mask R-CNN GitHub codes is applicable:
https://github.com/matterport/Mask_RCNN
The Keras source codes not only propose complex executable CNN structure but also give lots of useful utility for image processing.
Hello Marilyn Bello
I have used the Captom.ai library, designed for Model interpretability for PyTorch. It has many exciting features that you might find helpful as an explanation.
Here is a link to their project:
https://github.com/pytorch/captum
I hope this can help you. Best regards
Hello Marilyn bello
I sent you link that may be benefit you,
https://www.mdpi.com/2072-4292/12/18/3015/htm
- Badges
- Science topic