This model is based on two papers:
- Siamese Neural Networks for One-Shot Image Recognition, full thesis: Koch, Gregory R.. "Siamese Neural Networks for One-Shot Image Recognition" (2015) and ICML 2015 Accepted paper
- Squeeze-and-Excitation Networks
Aim is to gain better accuracy with modern SENet architecture, with same training pipeline as in 1-st paper.
Useful links:
- the most full implementation found on the web, with rich description: Goldesel23