Evaluating the faithfulness of saliency maps in explaining deep learning models using realistic perturbations.
José Pereira AmorimPedro H. AbreuJoão A. M. SantosMarc CortesVictor VilaPublished in: Inf. Process. Manag. (2023)
Keyphrases
- learning models
- saliency map
- visual attention
- input image
- visual saliency
- machine learning
- loss function
- semi supervised learning
- learning algorithm
- learning tasks
- machine learning algorithms
- human visual system
- natural images
- conditional random fields
- video frames
- learning problems
- visual features
- low level features
- bayesian networks
- dimensionality reduction
- classification models
- multimedia