Upload README.md with huggingface_hub

README.md

@@ -48,9 +48,7 @@ Five ground-roll intensity levels produce paired noisy / noise-label records:
 
 ## Model Architectures
 
-- **Attention UNet-Plus** (`atten_unet_plus`) — Wider Attention UNet variant. Base channels: 64, depth: 4.
-- **ResUNet-Plus** (`res_unet_plus`) — Wider ResUNet variant. Base channels: 64, depth: 4.
-- **UNet-Plus** (`unet_plus`) — Wider UNet variant. Base channels: 64, depth: 4.
+- **Enhanced Atten-UNet** (`enhanced_atten_unet`) — U-Net with residual blocks + attention-gated skip connections, trained with hybrid MSE + AFM (adaptive frequency modulation) loss. Predicts noise residual. Base channels: 32, depth: 4.
 
 ### Preprocessing