GRADCAM

Implementation of the Grad-CAM algorithm in an easy-to-use class, optimized for transfer learning projects and written using Keras and Tensorflow 2.x.

Requirement

• Python 3.6.x
• Matplotlib 3.4.x
• Numpy 1.19.x
• Tensorflow 2.4.x

Usage

Please take a look of examples folder for a complete example.

from gradcam import Gradcam

gc = Gradcam(model, 
             layer_name="top_conv",
             img_path=img_path,
             size=img_size,
             inner_model=model.get_layer("efficientnetb0"))

gc.generate_stack_img(save_name="../output/example_out")

Reference

[1] Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization.