wavedl 1.3.0__py3-none-any.whl → 1.4.0__py3-none-any.whl

wavedl/models/resnet3d.py

@@ -0,0 +1,258 @@
+"""
+ResNet3D: 3D Residual Networks for Volumetric Data
+===================================================
+
+3D extension of ResNet for processing volumetric data such as C-scans,
+3D wavefield imaging, and spatiotemporal cubes. Wraps torchvision's
+video models adapted for regression tasks.
+
+**Key Features**:
+    - Native 3D convolutions for volumetric processing
+    - Pretrained weights from Kinetics-400 (video action recognition)
+    - Adapted for single-channel input (grayscale volumes)
+    - Custom regression head for parameter estimation
+
+**Variants**:
+    - resnet3d_18: Lightweight (33M params)
+    - resnet3d_34: Medium depth
+    - resnet3d_50: Higher capacity with bottleneck blocks
+
+**Use Cases**:
+    - C-scan volume analysis (ultrasonic NDT)
+    - 3D wavefield imaging and inversion
+    - Spatiotemporal data cubes (time × space × space)
+    - Medical imaging (CT/MRI volumes)
+
+**Note**: ResNet3D is 3D-only. For 1D/2D data, use TCN or standard ResNet.
+
+References:
+    Hara, K., Kataoka, H., & Satoh, Y. (2018). Can Spatiotemporal 3D CNNs
+    Retrace the History of 2D CNNs and ImageNet? CVPR 2018.
+    https://arxiv.org/abs/1711.09577
+
+    He, K., et al. (2016). Deep Residual Learning for Image Recognition.
+    CVPR 2016. https://arxiv.org/abs/1512.03385
+
+Author: Ductho Le (ductho.le@outlook.com)
+"""
+
+from typing import Any
+
+import torch
+import torch.nn as nn
+
+
+try:
+    from torchvision.models.video import (
+        MC3_18_Weights,
+        R3D_18_Weights,
+        mc3_18,
+        r3d_18,
+    )
+
+    RESNET3D_AVAILABLE = True
+except ImportError:
+    RESNET3D_AVAILABLE = False
+
+from wavedl.models.base import BaseModel
+from wavedl.models.registry import register_model
+
+
+class ResNet3DBase(BaseModel):
+    """
+    Base ResNet3D class for volumetric regression tasks.
+
+    Wraps torchvision 3D ResNet with:
+    - Optional pretrained weights (Kinetics-400)
+    - Automatic input channel adaptation (grayscale → 3ch)
+    - Custom regression head
+
+    Note: This is 3D-only. Input shape must be (D, H, W).
+    """
+
+    def __init__(
+        self,
+        in_shape: tuple[int, int, int],
+        out_size: int,
+        model_fn,
+        weights_class,
+        pretrained: bool = True,
+        dropout_rate: float = 0.3,
+        freeze_backbone: bool = False,
+        regression_hidden: int = 512,
+        **kwargs,
+    ):
+        """
+        Initialize ResNet3D for regression.
+
+        Args:
+            in_shape: (D, H, W) input volume dimensions
+            out_size: Number of regression output targets
+            model_fn: torchvision model constructor
+            weights_class: Pretrained weights enum class
+            pretrained: Use Kinetics-400 pretrained weights (default: True)
+            dropout_rate: Dropout rate in regression head (default: 0.3)
+            freeze_backbone: Freeze backbone for fine-tuning (default: False)
+            regression_hidden: Hidden units in regression head (default: 512)
+        """
+        super().__init__(in_shape, out_size)
+
+        if not RESNET3D_AVAILABLE:
+            raise ImportError(
+                "torchvision >= 0.12 is required for ResNet3D. "
+                "Install with: pip install torchvision>=0.12"
+            )
+
+        if len(in_shape) != 3:
+            raise ValueError(
+                f"ResNet3D requires 3D input (D, H, W), got {len(in_shape)}D. "
+                "For 1D data, use TCN. For 2D data, use standard ResNet."
+            )
+
+        self.pretrained = pretrained
+        self.dropout_rate = dropout_rate
+        self.freeze_backbone = freeze_backbone
+        self.regression_hidden = regression_hidden
+
+        # Load pretrained backbone
+        weights = weights_class.DEFAULT if pretrained else None
+        self.backbone = model_fn(weights=weights)
+
+        # Get the fc input features
+        in_features = self.backbone.fc.in_features
+
+        # Replace fc with regression head
+        self.backbone.fc = nn.Sequential(
+            nn.Dropout(dropout_rate),
+            nn.Linear(in_features, regression_hidden),
+            nn.ReLU(inplace=True),
+            nn.Dropout(dropout_rate * 0.5),
+            nn.Linear(regression_hidden, regression_hidden // 2),
+            nn.ReLU(inplace=True),
+            nn.Linear(regression_hidden // 2, out_size),
+        )
+
+        # Optionally freeze backbone for fine-tuning
+        if freeze_backbone:
+            self._freeze_backbone()
+
+    def _freeze_backbone(self):
+        """Freeze all backbone parameters except the fc head."""
+        for name, param in self.backbone.named_parameters():
+            if "fc" not in name:
+                param.requires_grad = False
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Forward pass.
+
+        Args:
+            x: Input tensor of shape (B, C, D, H, W) where C is 1 or 3
+
+        Returns:
+            Output tensor of shape (B, out_size)
+        """
+        # Expand single channel to 3 channels for pretrained weights compatibility
+        if x.size(1) == 1:
+            x = x.expand(-1, 3, -1, -1, -1)
+
+        return self.backbone(x)
+
+    @classmethod
+    def get_default_config(cls) -> dict[str, Any]:
+        """Return default configuration for ResNet3D."""
+        return {
+            "pretrained": True,
+            "dropout_rate": 0.3,
+            "freeze_backbone": False,
+            "regression_hidden": 512,
+        }
+
+
+# =============================================================================
+# REGISTERED MODEL VARIANTS
+# =============================================================================
+
+
+@register_model("resnet3d_18")
+class ResNet3D18(ResNet3DBase):
+    """
+    ResNet3D-18: Lightweight 3D ResNet for volumetric data.
+
+    ~33M parameters. Uses 3D convolutions throughout for true volumetric processing.
+    Pretrained on Kinetics-400 (video action recognition).
+
+    Recommended for:
+        - C-scan ultrasonic inspection volumes
+        - 3D wavefield data cubes
+        - Medical imaging (CT/MRI)
+        - Moderate compute budgets
+
+    Args:
+        in_shape: (D, H, W) volume dimensions
+        out_size: Number of regression targets
+        pretrained: Use Kinetics-400 pretrained weights (default: True)
+        dropout_rate: Dropout rate in head (default: 0.3)
+        freeze_backbone: Freeze backbone for fine-tuning (default: False)
+        regression_hidden: Hidden units in regression head (default: 512)
+
+    Example:
+        >>> model = ResNet3D18(in_shape=(16, 112, 112), out_size=3)
+        >>> x = torch.randn(2, 1, 16, 112, 112)
+        >>> out = model(x)  # (2, 3)
+    """
+
+    def __init__(self, in_shape: tuple[int, int, int], out_size: int, **kwargs):
+        super().__init__(
+            in_shape=in_shape,
+            out_size=out_size,
+            model_fn=r3d_18,
+            weights_class=R3D_18_Weights,
+            **kwargs,
+        )
+
+    def __repr__(self) -> str:
+        pt = "pretrained" if self.pretrained else "scratch"
+        return f"ResNet3D_18({pt}, in={self.in_shape}, out={self.out_size})"
+
+
+@register_model("mc3_18")
+class MC3_18(ResNet3DBase):
+    """
+    MC3-18: Mixed Convolution 3D ResNet (3D stem + 2D residual blocks).
+
+    ~11M parameters. More efficient than pure 3D ResNet while maintaining
+    good spatiotemporal modeling. Uses 3D convolutions in early layers
+    and 2D convolutions in later layers.
+
+    Recommended for:
+        - When pure 3D is too expensive
+        - Volumes with limited temporal/depth extent
+        - Faster training with reasonable accuracy
+
+    Args:
+        in_shape: (D, H, W) volume dimensions
+        out_size: Number of regression targets
+        pretrained: Use Kinetics-400 pretrained weights (default: True)
+        dropout_rate: Dropout rate in head (default: 0.3)
+        freeze_backbone: Freeze backbone for fine-tuning (default: False)
+        regression_hidden: Hidden units in regression head (default: 512)
+
+    Example:
+        >>> model = MC3_18(in_shape=(16, 112, 112), out_size=3)
+        >>> x = torch.randn(2, 1, 16, 112, 112)
+        >>> out = model(x)  # (2, 3)
+    """
+
+    def __init__(self, in_shape: tuple[int, int, int], out_size: int, **kwargs):
+        super().__init__(
+            in_shape=in_shape,
+            out_size=out_size,
+            model_fn=mc3_18,
+            weights_class=MC3_18_Weights,
+            **kwargs,
+        )
+
+    def __repr__(self) -> str:
+        pt = "pretrained" if self.pretrained else "scratch"
+        return f"MC3_18({pt}, in={self.in_shape}, out={self.out_size})"

wavedl/models/swin.py

@@ -0,0 +1,390 @@
+"""
+Swin Transformer: Hierarchical Vision Transformer with Shifted Windows
+=======================================================================
+
+State-of-the-art vision transformer that computes self-attention within
+local windows while enabling cross-window connections via shifting.
+Achieves excellent accuracy with linear computational complexity.
+
+**Key Innovations**:
+    - Hierarchical feature maps (like CNNs) for multi-scale processing
+    - Shifted window attention: O(n) complexity vs O(n²) for vanilla ViT
+    - Local attention with global receptive field through layer stacking
+    - Strong inductive bias for structured data
+
+**Variants**:
+    - swin_t: Tiny (28M params) - Efficient default
+    - swin_s: Small (50M params) - Better accuracy
+    - swin_b: Base (88M params) - High accuracy
+
+**Why Swin over ViT?**:
+    - Better for smaller datasets (stronger inductive bias)
+    - Handles higher resolution inputs efficiently
+    - Produces hierarchical features (useful for multi-scale patterns)
+    - More efficient memory usage
+
+**Note**: Swin Transformer is 2D-only. For 1D data, use TCN. For 3D data, use ResNet3D.
+
+References:
+    Liu, Z., et al. (2021). Swin Transformer: Hierarchical Vision Transformer
+    using Shifted Windows. ICCV 2021 (Best Paper). https://arxiv.org/abs/2103.14030
+
+Author: Ductho Le (ductho.le@outlook.com)
+"""
+
+from typing import Any
+
+import torch
+import torch.nn as nn
+
+
+try:
+    from torchvision.models import (
+        Swin_B_Weights,
+        Swin_S_Weights,
+        Swin_T_Weights,
+        swin_b,
+        swin_s,
+        swin_t,
+    )
+
+    SWIN_AVAILABLE = True
+except ImportError:
+    SWIN_AVAILABLE = False
+
+from wavedl.models.base import BaseModel
+from wavedl.models.registry import register_model
+
+
+class SwinTransformerBase(BaseModel):
+    """
+    Base Swin Transformer class for regression tasks.
+
+    Wraps torchvision Swin Transformer with:
+    - Optional pretrained weights (ImageNet-1K or ImageNet-22K)
+    - Automatic input channel adaptation (grayscale → 3ch)
+    - Custom regression head with layer normalization
+
+    Swin Transformer excels at:
+    - Multi-scale feature extraction (dispersion curves, spectrograms)
+    - High-resolution inputs (efficient O(n) attention)
+    - Tasks requiring both local and global context
+    - Transfer learning from pretrained weights
+
+    Note: This is 2D-only. Input shape must be (H, W).
+    """
+
+    def __init__(
+        self,
+        in_shape: tuple[int, int],
+        out_size: int,
+        model_fn,
+        weights_class,
+        pretrained: bool = True,
+        dropout_rate: float = 0.3,
+        freeze_backbone: bool = False,
+        regression_hidden: int = 512,
+        **kwargs,
+    ):
+        """
+        Initialize Swin Transformer for regression.
+
+        Args:
+            in_shape: (H, W) input image dimensions
+            out_size: Number of regression output targets
+            model_fn: torchvision model constructor
+            weights_class: Pretrained weights enum class
+            pretrained: Use ImageNet pretrained weights (default: True)
+            dropout_rate: Dropout rate in regression head (default: 0.3)
+            freeze_backbone: Freeze backbone for fine-tuning (default: False)
+            regression_hidden: Hidden units in regression head (default: 512)
+        """
+        super().__init__(in_shape, out_size)
+
+        if not SWIN_AVAILABLE:
+            raise ImportError(
+                "torchvision >= 0.12 is required for Swin Transformer. "
+                "Install with: pip install torchvision>=0.12"
+            )
+
+        if len(in_shape) != 2:
+            raise ValueError(
+                f"Swin Transformer requires 2D input (H, W), got {len(in_shape)}D. "
+                "For 1D data, use TCN. For 3D data, use ResNet3D."
+            )
+
+        self.pretrained = pretrained
+        self.dropout_rate = dropout_rate
+        self.freeze_backbone = freeze_backbone
+        self.regression_hidden = regression_hidden
+
+        # Load pretrained backbone
+        weights = weights_class.IMAGENET1K_V1 if pretrained else None
+        self.backbone = model_fn(weights=weights)
+
+        # Swin Transformer head structure:
+        # head: Linear (embed_dim → num_classes)
+        # We need to get the embedding dimension from the head
+
+        in_features = self.backbone.head.in_features
+
+        # Replace head with regression head
+        # Use LayerNorm for stability (matches Transformer architecture)
+        self.backbone.head = nn.Sequential(
+            nn.LayerNorm(in_features),
+            nn.Dropout(dropout_rate),
+            nn.Linear(in_features, regression_hidden),
+            nn.GELU(),  # GELU matches Transformer's activation
+            nn.Dropout(dropout_rate * 0.5),
+            nn.Linear(regression_hidden, regression_hidden // 2),
+            nn.GELU(),
+            nn.Linear(regression_hidden // 2, out_size),
+        )
+
+        # Optionally freeze backbone for fine-tuning
+        if freeze_backbone:
+            self._freeze_backbone()
+
+    def _freeze_backbone(self):
+        """Freeze all backbone parameters except the head."""
+        for name, param in self.backbone.named_parameters():
+            if "head" not in name:
+                param.requires_grad = False
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Forward pass.
+
+        Args:
+            x: Input tensor of shape (B, C, H, W) where C is 1 or 3
+
+        Returns:
+            Output tensor of shape (B, out_size)
+        """
+        # Expand single channel to 3 channels for pretrained weights compatibility
+        if x.size(1) == 1:
+            x = x.expand(-1, 3, -1, -1)
+
+        return self.backbone(x)
+
+    @classmethod
+    def get_default_config(cls) -> dict[str, Any]:
+        """Return default configuration for Swin Transformer."""
+        return {
+            "pretrained": True,
+            "dropout_rate": 0.3,
+            "freeze_backbone": False,
+            "regression_hidden": 512,
+        }
+
+    def get_optimizer_groups(self, base_lr: float, weight_decay: float = 0.05) -> list:
+        """
+        Get parameter groups with layer-wise learning rate decay.
+
+        Swin Transformer benefits from decaying learning rate for earlier layers.
+        This is a common practice for fine-tuning vision transformers.
+
+        Args:
+            base_lr: Base learning rate (applied to head)
+            weight_decay: Weight decay coefficient
+
+        Returns:
+            List of parameter group dictionaries
+        """
+        # Separate parameters: head (full LR) vs backbone (decayed LR)
+        head_params = []
+        backbone_params = []
+        no_decay_params = []
+
+        for name, param in self.backbone.named_parameters():
+            if not param.requires_grad:
+                continue
+
+            # No weight decay for bias and normalization
+            if "bias" in name or "norm" in name:
+                no_decay_params.append(param)
+            elif "head" in name:
+                head_params.append(param)
+            else:
+                backbone_params.append(param)
+
+        groups = []
+
+        if head_params:
+            groups.append(
+                {
+                    "params": head_params,
+                    "lr": base_lr,
+                    "weight_decay": weight_decay,
+                }
+            )
+
+        if backbone_params:
+            # Apply 0.1x learning rate to backbone (common for fine-tuning)
+            groups.append(
+                {
+                    "params": backbone_params,
+                    "lr": base_lr * 0.1,
+                    "weight_decay": weight_decay,
+                }
+            )
+
+        if no_decay_params:
+            groups.append(
+                {
+                    "params": no_decay_params,
+                    "lr": base_lr,
+                    "weight_decay": 0.0,
+                }
+            )
+
+        return groups if groups else [{"params": self.parameters(), "lr": base_lr}]
+
+
+# =============================================================================
+# REGISTERED MODEL VARIANTS
+# =============================================================================
+
+
+@register_model("swin_t")
+class SwinTiny(SwinTransformerBase):
+    """
+    Swin-T (Tiny): Efficient default for most wave-based tasks.
+
+    ~28M parameters. Good balance of accuracy and computational cost.
+    Outperforms ResNet50 while being more efficient.
+
+    Recommended for:
+        - Default choice for 2D wave data
+        - Dispersion curves, spectrograms, B-scans
+        - When hierarchical features matter
+        - Transfer learning with limited data
+
+    Architecture:
+        - Patch size: 4×4
+        - Window size: 7×7
+        - Embed dim: 96
+        - Depths: [2, 2, 6, 2]
+        - Heads: [3, 6, 12, 24]
+
+    Args:
+        in_shape: (H, W) image dimensions
+        out_size: Number of regression targets
+        pretrained: Use ImageNet pretrained weights (default: True)
+        dropout_rate: Dropout rate in head (default: 0.3)
+        freeze_backbone: Freeze backbone for fine-tuning (default: False)
+        regression_hidden: Hidden units in regression head (default: 512)
+
+    Example:
+        >>> model = SwinTiny(in_shape=(224, 224), out_size=3)
+        >>> x = torch.randn(4, 1, 224, 224)
+        >>> out = model(x)  # (4, 3)
+    """
+
+    def __init__(self, in_shape: tuple[int, int], out_size: int, **kwargs):
+        super().__init__(
+            in_shape=in_shape,
+            out_size=out_size,
+            model_fn=swin_t,
+            weights_class=Swin_T_Weights,
+            **kwargs,
+        )
+
+    def __repr__(self) -> str:
+        pt = "pretrained" if self.pretrained else "scratch"
+        return f"Swin_Tiny({pt}, in={self.in_shape}, out={self.out_size})"
+
+
+@register_model("swin_s")
+class SwinSmall(SwinTransformerBase):
+    """
+    Swin-S (Small): Higher accuracy with moderate compute.
+
+    ~50M parameters. Better accuracy than Swin-T for larger datasets.
+
+    Recommended for:
+        - Larger datasets (>20k samples)
+        - When Swin-T doesn't provide enough capacity
+        - Complex multi-scale patterns
+
+    Architecture:
+        - Patch size: 4×4
+        - Window size: 7×7
+        - Embed dim: 96
+        - Depths: [2, 2, 18, 2]  (deeper stage 3)
+        - Heads: [3, 6, 12, 24]
+
+    Args:
+        in_shape: (H, W) image dimensions
+        out_size: Number of regression targets
+        pretrained: Use ImageNet pretrained weights (default: True)
+        dropout_rate: Dropout rate in head (default: 0.3)
+        freeze_backbone: Freeze backbone for fine-tuning (default: False)
+        regression_hidden: Hidden units in regression head (default: 512)
+
+    Example:
+        >>> model = SwinSmall(in_shape=(224, 224), out_size=3)
+        >>> x = torch.randn(4, 1, 224, 224)
+        >>> out = model(x)  # (4, 3)
+    """
+
+    def __init__(self, in_shape: tuple[int, int], out_size: int, **kwargs):
+        super().__init__(
+            in_shape=in_shape,
+            out_size=out_size,
+            model_fn=swin_s,
+            weights_class=Swin_S_Weights,
+            **kwargs,
+        )
+
+    def __repr__(self) -> str:
+        pt = "pretrained" if self.pretrained else "scratch"
+        return f"Swin_Small({pt}, in={self.in_shape}, out={self.out_size})"
+
+
+@register_model("swin_b")
+class SwinBase(SwinTransformerBase):
+    """
+    Swin-B (Base): Maximum accuracy for large-scale tasks.
+
+    ~88M parameters. Best accuracy but requires more compute and data.
+
+    Recommended for:
+        - Very large datasets (>50k samples)
+        - When accuracy is more important than efficiency
+        - HPC environments with ample GPU memory
+        - Research experiments
+
+    Architecture:
+        - Patch size: 4×4
+        - Window size: 7×7
+        - Embed dim: 128
+        - Depths: [2, 2, 18, 2]
+        - Heads: [4, 8, 16, 32]
+
+    Args:
+        in_shape: (H, W) image dimensions
+        out_size: Number of regression targets
+        pretrained: Use ImageNet pretrained weights (default: True)
+        dropout_rate: Dropout rate in head (default: 0.3)
+        freeze_backbone: Freeze backbone for fine-tuning (default: False)
+        regression_hidden: Hidden units in regression head (default: 512)
+
+    Example:
+        >>> model = SwinBase(in_shape=(224, 224), out_size=3)
+        >>> x = torch.randn(4, 1, 224, 224)
+        >>> out = model(x)  # (4, 3)
+    """
+
+    def __init__(self, in_shape: tuple[int, int], out_size: int, **kwargs):
+        super().__init__(
+            in_shape=in_shape,
+            out_size=out_size,
+            model_fn=swin_b,
+            weights_class=Swin_B_Weights,
+            **kwargs,
+        )
+
+    def __repr__(self) -> str:
+        pt = "pretrained" if self.pretrained else "scratch"
+        return f"Swin_Base({pt}, in={self.in_shape}, out={self.out_size})"