model.py

import torch
from monai.networks.nets import DynUNet
import os

def load_model(model_path="best_model_large_data.pth", device="cpu"):
    """Load DynUNet model with weights"""
    try:
        model = DynUNet(
            spatial_dims=2,
            in_channels=1,
            out_channels=1,
            kernel_size=[3, 3, 3, 3, 3],
            strides=[1, 2, 2, 2, 2],
            upsample_kernel_size=[2, 2, 2, 2],
            filters=[32, 64, 128, 256, 512],
            norm_name="INSTANCE",
            res_block=True,
            deep_supervision=False,
        )
        
        state_dict = torch.load(model_path, map_location=device)
        model.load_state_dict(state_dict)
        model.to(device)
        model.eval()
        return model

    except Exception as e:
        print(f"❌ Model initialization failed: {e}")
        raise

def predict_mask(model, image_tensor):
    """Predict segmentation mask with sigmoid activation."""
    try:
        if image_tensor.dim() != 4 or image_tensor.shape[1] != 1:
            raise ValueError(f"Input tensor must be [1, 1, H, W]. Got {image_tensor.shape}")
        
        with torch.no_grad():
            return torch.sigmoid(model(image_tensor))
            
    except Exception as e:
        print(f"Prediction failed: {e}")
        raise