Table 1. Pixel-level masked D-MNet... | Emerald Publishing

Table 1.

Pixel-level masked D-MNet transformer setting

Training setup		Model parameters
Seed	1	Loss	L1 and MSE
Patch size	96	Self-attention heads	8
Epochs	1000	Optimization	Adam
Learning rate	2 × 10⁻⁴	D-MNet layer	9