Visualization of Mask RCNN classification focus areas using methods like GradCAM

matterport / Mask_RCNN

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow

Other

24.52k stars 11.68k forks source link

Visualization of Mask RCNN classification focus areas using methods like GradCAM #1984

Open mist-wing2020 opened 4 years ago

mist-wing2020 commented 4 years ago

Hi,

During my classification process, I would like to visualize the focus areas of the Mask RCNN neural network.

I am aware of methods like GradCAM that enable the same. I tried implementing the GradCAM basics to the mrcnn_class_logits layer, but this gave me an incorrect output.

Is there any way to visualize where the neural network is focusing before it assigns the class to a particular instance?

Any help or guidance will be appreciated.

mminakshi commented 4 years ago

Hi mist-wing2020 Could you please let me know how you implemented GradCAM here.

Trotts commented 4 years ago

Hi,

I'm also wanting to implement Grad-CAM on a trained Mask-RCNN. Did anyone have any luck with this, @mist-wing2020 @mminakshi ?

FloBr49 commented 3 years ago

Hi @mist-wing2020 @mminakshi @Trotts,

has anybody figured out how to implement GradCAM? Thank you.

Kartiky246 commented 2 years ago

Hi @FloBr49 @Trotts @mminakshi @mist-wing2020 has anybody figured out how to implement GradCAM? Thank you.