A novel feature-scrambling approach reveals the capacity of convolutional neural networks to learn spatial relations.
Amr FarahatFelix EffenbergerMartin A. VinckPublished in: Neural Networks (2023)
Keyphrases
- spatial relations
- convolutional neural networks
- spatial reasoning
- spatial information
- spatial configuration
- spatial relationships
- segmented regions
- closely related
- spatial objects
- topological relations
- directional relations
- temporal relations
- image features
- visual words
- spatial knowledge
- feature vectors
- physical objects
- data mining
- convolutional network