/Apply-XAI-on-ResNet50

Generate explanations for the ResNet50 classification using Grad-CAM and LIME (XAI Method)

Primary LanguageJupyter Notebook

Implementation of XAI method on ResNet50

Description

Course: Machine Learning (Fall 2022)
Task: Generate explanations for the classification results by ResNet50 using Grad-CAM and LIME. Implement with https://keras.io/examples/vision/grad_cam/ and https://github.com/marcotcr/lime

Pre-trained model that was used:

  • ResNet50 Model

Explanation AI used:

  • Grad-CAM
  • LIME

Analysis:

The Grad-CAM highlights the localized areas on the image that contributes positively to the model’s classification.
The LIME explanation method visualizes the contribution of each feature of an image to the model’s classification.
Conclusion, that the ResNet50 model can correctly identify and focus on the right areas of the image to make its classification