An implementation of Score-CAM with keras (tf.keras).
The following methods are compared.
- Grad-CAM
- Grad-CAM++
- Score-CAM
- Faster-Score-CAM (faster version of Score-CAM)
In addition, the followings are contained.
- Visualization of anomalous area for DAGM dataset
- Sample code for applying Score-CAM to your model.
Blog post: Qiita
We thought that several channels were dominant in generating the final heat map. Faster-Score-CAM adds the processing of “use only channels with large variances as mask images” to Score-CAM. (max_N = -1 is the original Score-CAM).
When using VGG16, Faster-Score-CAM is about 10 times faster than Score-CAM.
- Python >= 3.6
- tensorflow >= 2.0
See Score-CAM.ipynb.