Swin Transformer (Swin-T)
This is pretty much state of the art
There have a few variants:
- Swin-T: C = 96, layer numbers = {2, 2, 6, 2}
- Swin-S: C = 96, layer numbers ={2, 2, 18, 2}
- Swin-B: C = 128, layer numbers ={2, 2, 18, 2}
- Swin-L: C = 192, layer numbers ={2, 2, 18, 2}
Citation
(Liu et al., 2021)
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021). Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. https://arxiv.org/pdf/2103.14030