Upload models/hybrid_model.py with huggingface_hub
Browse files- models/hybrid_model.py +211 -0
models/hybrid_model.py
ADDED
|
@@ -0,0 +1,211 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Hybrid CNN-ViT Food Classifier
|
| 3 |
+
Combines ResNet50 and DeiT-Base with adaptive fusion
|
| 4 |
+
"""
|
| 5 |
+
import torch
|
| 6 |
+
import torch.nn as nn
|
| 7 |
+
import torch.nn.functional as F
|
| 8 |
+
from typing import Dict, Any, Optional
|
| 9 |
+
|
| 10 |
+
from .cnn_branch import CNNBranch
|
| 11 |
+
from .vit_branch import ViTBranch
|
| 12 |
+
from .fusion_module import AdaptiveFusionModule
|
| 13 |
+
|
| 14 |
+
class HybridFoodClassifier(nn.Module):
|
| 15 |
+
"""Hybrid CNN-ViT model for food classification"""
|
| 16 |
+
|
| 17 |
+
def __init__(
|
| 18 |
+
self,
|
| 19 |
+
num_classes: int = 101,
|
| 20 |
+
feature_dim: int = 768,
|
| 21 |
+
hidden_dim: int = 512,
|
| 22 |
+
dropout: float = 0.2,
|
| 23 |
+
pretrained: bool = True,
|
| 24 |
+
freeze_early_layers: bool = True
|
| 25 |
+
):
|
| 26 |
+
super(HybridFoodClassifier, self).__init__()
|
| 27 |
+
|
| 28 |
+
self.num_classes = num_classes
|
| 29 |
+
self.feature_dim = feature_dim
|
| 30 |
+
self.hidden_dim = hidden_dim
|
| 31 |
+
|
| 32 |
+
# CNN Branch (ResNet50)
|
| 33 |
+
self.cnn_branch = CNNBranch(
|
| 34 |
+
pretrained=pretrained,
|
| 35 |
+
freeze_early_layers=freeze_early_layers,
|
| 36 |
+
dropout=dropout,
|
| 37 |
+
feature_dim=feature_dim
|
| 38 |
+
)
|
| 39 |
+
|
| 40 |
+
# ViT Branch (DeiT-Base)
|
| 41 |
+
self.vit_branch = ViTBranch(
|
| 42 |
+
pretrained=pretrained,
|
| 43 |
+
freeze_early_layers=freeze_early_layers,
|
| 44 |
+
dropout=dropout,
|
| 45 |
+
feature_dim=feature_dim
|
| 46 |
+
)
|
| 47 |
+
|
| 48 |
+
# Fusion Module
|
| 49 |
+
self.fusion_module = AdaptiveFusionModule(
|
| 50 |
+
feature_dim=feature_dim,
|
| 51 |
+
hidden_dim=hidden_dim,
|
| 52 |
+
dropout=dropout
|
| 53 |
+
)
|
| 54 |
+
|
| 55 |
+
# Classification Head
|
| 56 |
+
self.classifier = nn.Sequential(
|
| 57 |
+
nn.Linear(hidden_dim, hidden_dim // 2),
|
| 58 |
+
nn.LayerNorm(hidden_dim // 2),
|
| 59 |
+
nn.GELU(),
|
| 60 |
+
nn.Dropout(dropout),
|
| 61 |
+
nn.Linear(hidden_dim // 2, num_classes)
|
| 62 |
+
)
|
| 63 |
+
|
| 64 |
+
# Auxiliary classifiers for training stability
|
| 65 |
+
self.cnn_aux_classifier = nn.Sequential(
|
| 66 |
+
nn.Linear(feature_dim, hidden_dim // 2),
|
| 67 |
+
nn.ReLU(),
|
| 68 |
+
nn.Dropout(dropout),
|
| 69 |
+
nn.Linear(hidden_dim // 2, num_classes)
|
| 70 |
+
)
|
| 71 |
+
|
| 72 |
+
self.vit_aux_classifier = nn.Sequential(
|
| 73 |
+
nn.Linear(feature_dim, hidden_dim // 2),
|
| 74 |
+
nn.ReLU(),
|
| 75 |
+
nn.Dropout(dropout),
|
| 76 |
+
nn.Linear(hidden_dim // 2, num_classes)
|
| 77 |
+
)
|
| 78 |
+
|
| 79 |
+
# Initialize weights
|
| 80 |
+
self._initialize_weights()
|
| 81 |
+
|
| 82 |
+
def _initialize_weights(self):
|
| 83 |
+
"""Initialize classifier weights"""
|
| 84 |
+
for m in [self.classifier, self.cnn_aux_classifier, self.vit_aux_classifier]:
|
| 85 |
+
for layer in m:
|
| 86 |
+
if isinstance(layer, nn.Linear):
|
| 87 |
+
nn.init.xavier_uniform_(layer.weight)
|
| 88 |
+
if layer.bias is not None:
|
| 89 |
+
nn.init.constant_(layer.bias, 0)
|
| 90 |
+
|
| 91 |
+
def forward(
|
| 92 |
+
self,
|
| 93 |
+
x: torch.Tensor,
|
| 94 |
+
return_features: bool = False,
|
| 95 |
+
use_aux_loss: bool = True
|
| 96 |
+
) -> Dict[str, torch.Tensor]:
|
| 97 |
+
"""
|
| 98 |
+
Forward pass
|
| 99 |
+
|
| 100 |
+
Args:
|
| 101 |
+
x: Input tensor [B, 3, H, W]
|
| 102 |
+
return_features: Whether to return intermediate features
|
| 103 |
+
use_aux_loss: Whether to compute auxiliary losses
|
| 104 |
+
|
| 105 |
+
Returns:
|
| 106 |
+
Dictionary containing logits and optionally features/aux_logits
|
| 107 |
+
"""
|
| 108 |
+
# CNN Branch
|
| 109 |
+
cnn_spatial, cnn_global = self.cnn_branch(x)
|
| 110 |
+
|
| 111 |
+
# ViT Branch
|
| 112 |
+
vit_spatial, vit_global = self.vit_branch(x)
|
| 113 |
+
|
| 114 |
+
# Fusion
|
| 115 |
+
fused_spatial, fused_global = self.fusion_module(
|
| 116 |
+
cnn_spatial, cnn_global, vit_spatial, vit_global
|
| 117 |
+
)
|
| 118 |
+
|
| 119 |
+
# Main classification
|
| 120 |
+
logits = self.classifier(fused_global)
|
| 121 |
+
|
| 122 |
+
# Prepare output
|
| 123 |
+
output = {'logits': logits}
|
| 124 |
+
|
| 125 |
+
# Auxiliary losses for training
|
| 126 |
+
if use_aux_loss and self.training:
|
| 127 |
+
cnn_aux_logits = self.cnn_aux_classifier(cnn_global)
|
| 128 |
+
vit_aux_logits = self.vit_aux_classifier(vit_global)
|
| 129 |
+
output.update({
|
| 130 |
+
'cnn_aux_logits': cnn_aux_logits,
|
| 131 |
+
'vit_aux_logits': vit_aux_logits
|
| 132 |
+
})
|
| 133 |
+
|
| 134 |
+
# Return features if requested
|
| 135 |
+
if return_features:
|
| 136 |
+
output.update({
|
| 137 |
+
'cnn_spatial': cnn_spatial,
|
| 138 |
+
'cnn_global': cnn_global,
|
| 139 |
+
'vit_spatial': vit_spatial,
|
| 140 |
+
'vit_global': vit_global,
|
| 141 |
+
'fused_spatial': fused_spatial,
|
| 142 |
+
'fused_global': fused_global
|
| 143 |
+
})
|
| 144 |
+
|
| 145 |
+
return output
|
| 146 |
+
|
| 147 |
+
def get_attention_maps(self, x: torch.Tensor) -> Dict[str, torch.Tensor]:
|
| 148 |
+
"""Get attention maps for visualization"""
|
| 149 |
+
with torch.no_grad():
|
| 150 |
+
# Get features
|
| 151 |
+
output = self.forward(x, return_features=True, use_aux_loss=False)
|
| 152 |
+
|
| 153 |
+
# CNN attention (using global average pooling weights)
|
| 154 |
+
cnn_spatial = output['cnn_spatial'] # [B, feature_dim, 7, 7]
|
| 155 |
+
cnn_attention = torch.mean(cnn_spatial, dim=1, keepdim=True) # [B, 1, 7, 7]
|
| 156 |
+
cnn_attention = F.interpolate(
|
| 157 |
+
cnn_attention,
|
| 158 |
+
size=(224, 224),
|
| 159 |
+
mode='bilinear',
|
| 160 |
+
align_corners=False
|
| 161 |
+
) # [B, 1, 224, 224]
|
| 162 |
+
|
| 163 |
+
# ViT attention (using patch importance)
|
| 164 |
+
vit_spatial = output['vit_spatial'] # [B, 197, feature_dim] (196 patches + 1 CLS)
|
| 165 |
+
vit_patches = vit_spatial[:, 1:] # Remove CLS token, get [B, 196, feature_dim]
|
| 166 |
+
vit_attention = torch.mean(vit_patches, dim=-1) # [B, 196]
|
| 167 |
+
vit_attention = vit_attention.view(-1, 14, 14).unsqueeze(1) # [B, 1, 14, 14]
|
| 168 |
+
vit_attention = F.interpolate(
|
| 169 |
+
vit_attention,
|
| 170 |
+
size=(224, 224),
|
| 171 |
+
mode='bilinear',
|
| 172 |
+
align_corners=False
|
| 173 |
+
) # [B, 1, 224, 224]
|
| 174 |
+
|
| 175 |
+
return {
|
| 176 |
+
'cnn_attention': cnn_attention,
|
| 177 |
+
'vit_attention': vit_attention
|
| 178 |
+
}
|
| 179 |
+
|
| 180 |
+
def freeze_backbone(self):
|
| 181 |
+
"""Freeze backbone networks"""
|
| 182 |
+
for param in self.cnn_branch.backbone.parameters():
|
| 183 |
+
param.requires_grad = False
|
| 184 |
+
for param in self.vit_branch.vit.parameters():
|
| 185 |
+
param.requires_grad = False
|
| 186 |
+
|
| 187 |
+
def unfreeze_backbone(self):
|
| 188 |
+
"""Unfreeze backbone networks"""
|
| 189 |
+
for param in self.cnn_branch.backbone.parameters():
|
| 190 |
+
param.requires_grad = True
|
| 191 |
+
for param in self.vit_branch.vit.parameters():
|
| 192 |
+
param.requires_grad = True
|
| 193 |
+
|
| 194 |
+
def get_model_size(self) -> Dict[str, int]:
|
| 195 |
+
"""Get model size information"""
|
| 196 |
+
total_params = sum(p.numel() for p in self.parameters())
|
| 197 |
+
trainable_params = sum(p.numel() for p in self.parameters() if p.requires_grad)
|
| 198 |
+
|
| 199 |
+
cnn_params = sum(p.numel() for p in self.cnn_branch.parameters())
|
| 200 |
+
vit_params = sum(p.numel() for p in self.vit_branch.parameters())
|
| 201 |
+
fusion_params = sum(p.numel() for p in self.fusion_module.parameters())
|
| 202 |
+
classifier_params = sum(p.numel() for p in self.classifier.parameters())
|
| 203 |
+
|
| 204 |
+
return {
|
| 205 |
+
'total_params': total_params,
|
| 206 |
+
'trainable_params': trainable_params,
|
| 207 |
+
'cnn_params': cnn_params,
|
| 208 |
+
'vit_params': vit_params,
|
| 209 |
+
'fusion_params': fusion_params,
|
| 210 |
+
'classifier_params': classifier_params
|
| 211 |
+
}
|