A Long Skip Connection for Enhanced Color Selectivity in CNN Architectures
Some recent studies show that filters in convolutional neural networks (CNNs) have low color selectivity in datasets of natural scenes such as Imagenet. CNNs, bio-inspired by the visual cortex, are characterized by their hierarchical learning structure which appears to gradually transform the repres...
| Autores principales: | , , , |
|---|---|
| Formato: | Artículo |
| Lenguaje: | English |
| Publicado: |
MDPI AG
2023-08-01
|
| Colección: | Sensors |
| Materias: | |
| Acceso en línea: | https://www.mdpi.com/1424-8220/23/17/7582 |