Merge branch 'dev_mobilenetv2' into dev_shufflenetv1

mmcls/models/backbones/mobilenet_v2.py

@@ -2,8 +2,8 @@ import logging
 
 import torch.nn as nn
 import torch.utils.checkpoint as cp
+from mmcv.runner import load_checkpoint
 
-from ..runner import load_checkpoint
 from .base_backbone import BaseBackbone
 from .weight_init import constant_init, kaiming_init
 
@@ -22,13 +22,12 @@ def conv3x3(in_planes, out_planes, stride=1, dilation=1):
 
 def conv_1x1_bn(inp, oup, activation=nn.ReLU6):
     return nn.Sequential(
-        nn.Conv2d(inp, oup, 1, 1, 0, bias=False),
+        nn.Conv2d(inp, oup, 1, 1, 0, bias=False), nn.BatchNorm2d(oup),
-        nn.BatchNorm2d(oup),
+        activation(inplace=True))
-        activation(inplace=True)
-    )
 
 
 class ConvBNReLU(nn.Sequential):
+
     def __init__(self,
                  in_planes,
                  out_planes,
@@ -39,16 +38,15 @@ class ConvBNReLU(nn.Sequential):
         padding = (kernel_size - 1) // 2
 
         super(ConvBNReLU, self).__init__(
-            nn.Conv2d(in_planes,
+            nn.Conv2d(
-                      out_planes,
+                in_planes,
-                      kernel_size,
+                out_planes,
-                      stride,
+                kernel_size,
-                      padding,
+                stride,
-                      groups=groups,
+                padding,
-                      bias=False),
+                groups=groups,
-            nn.BatchNorm2d(out_planes),
+                bias=False), nn.BatchNorm2d(out_planes),
-            activation(inplace=True)
+            activation(inplace=True))
-        )
 
 
 def _make_divisible(v, divisor, min_value=None):
@@ -62,6 +60,7 @@ def _make_divisible(v, divisor, min_value=None):
 
 
 class InvertedResidual(nn.Module):
+
     def __init__(self,
                  inplanes,
                  outplanes,
@@ -79,17 +78,18 @@ class InvertedResidual(nn.Module):
         layers = []
         if expand_ratio != 1:
             # pw
-            layers.append(ConvBNReLU(inplanes,
+            layers.append(
-                                     hidden_dim,
+                ConvBNReLU(
-                                     kernel_size=1,
+                    inplanes, hidden_dim, kernel_size=1,
-                                     activation=activation))
+                    activation=activation))
         layers.extend([
             # dw
-            ConvBNReLU(hidden_dim,
+            ConvBNReLU(
-                       hidden_dim,
+                hidden_dim,
-                       stride=stride,
+                hidden_dim,
-                       groups=hidden_dim,
+                stride=stride,
-                       activation=activation),
+                groups=hidden_dim,
+                activation=activation),
             # pw-linear
             nn.Conv2d(hidden_dim, outplanes, 1, 1, 0, bias=False),
             nn.BatchNorm2d(outplanes),
@@ -97,6 +97,7 @@ class InvertedResidual(nn.Module):
         self.conv = nn.Sequential(*layers)
 
     def forward(self, x):
+
         def _inner_forward(x):
             if self.use_res_connect:
                 return x + self.conv(x)
@@ -122,15 +123,23 @@ def make_inverted_res_layer(block,
     layers = []
     for i in range(num_blocks):
         if i == 0:
-            layers.append(block(inplanes, planes, stride,
+            layers.append(
-                                expand_ratio=expand_ratio,
+                block(
-                                activation=activation,
+                    inplanes,
-                                with_cp=with_cp))
+                    planes,
+                    stride,
+                    expand_ratio=expand_ratio,
+                    activation=activation,
+                    with_cp=with_cp))
         else:
-            layers.append(block(inplanes, planes, 1,
+            layers.append(
-                                expand_ratio=expand_ratio,
+                block(
-                                activation=activation,
+                    inplanes,
-                                with_cp=with_cp))
+                    planes,
+                    1,
+                    expand_ratio=expand_ratio,
+                    activation=activation,
+                    with_cp=with_cp))
         inplanes = planes
     return nn.Sequential(*layers)
 
@@ -154,7 +163,7 @@ class MobileNetv2(BaseBackbone):
     def __init__(self,
                  widen_factor=1.,
                  activation=nn.ReLU6,
-                 out_indices=(0, 1, 2, 3, 4, 5, 6),
+                 out_indices=(0, 1, 2, 3, 4, 5, 6, 7),
                  frozen_stages=-1,
                  bn_eval=True,
                  bn_frozen=False,
@@ -162,21 +171,17 @@ class MobileNetv2(BaseBackbone):
         super(MobileNetv2, self).__init__()
         block = InvertedResidual
         # expand_ratio, out_channel, n, stride
-        inverted_residual_setting = [
+        inverted_residual_setting = [[1, 16, 1, 1], [6, 24, 2,
-            [1, 16, 1, 1],
+                                                     2], [6, 32, 3, 2],
-            [6, 24, 2, 2],
+                                     [6, 64, 4, 2], [6, 96, 3, 1],
-            [6, 32, 3, 2],
+                                     [6, 160, 3, 2], [6, 320, 1, 1]]
-            [6, 64, 4, 2],
-            [6, 96, 3, 1],
-            [6, 160, 3, 2],
-            [6, 320, 1, 1]
-        ]
         self.widen_factor = widen_factor
         if isinstance(activation, str):
             activation = eval(activation)
         self.activation = activation(inplace=True)
 
         self.out_indices = out_indices
+        assert frozen_stages <= 7
         self.frozen_stages = frozen_stages
         self.bn_eval = bn_eval
         self.bn_frozen = bn_frozen
@@ -210,9 +215,8 @@ class MobileNetv2(BaseBackbone):
         self.out_channel = int(self.out_channel * widen_factor) \
             if widen_factor > 1.0 else self.out_channel
 
-        self.conv_last = nn.Conv2d(self.inplanes,
+        self.conv_last = nn.Conv2d(
-                                   self.out_channel,
+            self.inplanes, self.out_channel, 1, 1, 0, bias=False)
-                                   1, 1, 0, bias=False)
         self.bn_last = nn.BatchNorm2d(self.out_channel)
 
         self.feat_dim = self.out_channel

tests/test_backbone.py

@@ -1,11 +1,118 @@
+import pytest
 import torch
 import torch.nn as nn
+from torch.nn.modules import GroupNorm
+from torch.nn.modules.batchnorm import _BatchNorm
 
 from mmcls.models.backbones import MobileNetv2
+from mmcls.models.backbones.mobilenet_v2 import InvertedResidual
+
+
+def is_block(modules):
+    """Check if is ResNet building block."""
+    if isinstance(modules, (InvertedResidual, )):
+        return True
+    return False
+
+
+def is_norm(modules):
+    """Check if is one of the norms."""
+    if isinstance(modules, (GroupNorm, _BatchNorm)):
+        return True
+    return False
+
+
+def check_norm_state(modules, train_state):
+    """Check if norm layer is in correct train state."""
+    for mod in modules:
+        if isinstance(mod, _BatchNorm):
+            if mod.training != train_state:
+                return False
+    return True
+
+
+def test_mobilenetv2_invertedresidual():
+
+    with pytest.raises(AssertionError):
+        # stride must be in [1, 2]
+        InvertedResidual(64, 16, stride=3, expand_ratio=6)
+
+    # Test InvertedResidual with checkpoint forward, stride=1
+    block = InvertedResidual(64, 16, stride=1, expand_ratio=6)
+    x = torch.randn(1, 64, 56, 56)
+    x_out = block(x)
+    assert x_out.shape == torch.Size([1, 16, 56, 56])
+
+    # Test InvertedResidual with checkpoint forward, stride=2
+    block = InvertedResidual(64, 16, stride=2, expand_ratio=6)
+    x = torch.randn(1, 64, 56, 56)
+    x_out = block(x)
+    assert x_out.shape == torch.Size([1, 16, 28, 28])
+
+    # Test InvertedResidual with checkpoint forward
+    block = InvertedResidual(64, 16, stride=1, expand_ratio=6, with_cp=True)
+    assert block.with_cp
+    x = torch.randn(1, 64, 56, 56)
+    x_out = block(x)
+    assert x_out.shape == torch.Size([1, 16, 56, 56])
+
+    # Test InvertedResidual with activation=nn.ReLU
+    block = InvertedResidual(
+        64, 16, stride=1, expand_ratio=6, activation=nn.ReLU)
+    x = torch.randn(1, 64, 56, 56)
+    x_out = block(x)
+    assert x_out.shape == torch.Size([1, 16, 56, 56])
 
 
 def test_mobilenetv2_backbone():
-    # Test MobileNetv2 with widen_factor 1.0, activation nn.ReLU6
+    with pytest.raises(TypeError):
+        # pretrained must be a string path
+        model = MobileNetv2()
+        model.init_weights(pretrained=0)
+
+    with pytest.raises(AssertionError):
+        # frozen_stages must less than 7
+        MobileNetv2(frozen_stages=8)
+
+    # Test MobileNetv2
+    model = MobileNetv2()
+    model.init_weights()
+    model.train()
+    assert check_norm_state(model.modules(), False)
+
+    # Test MobileNetv2 with first stage frozen
+    frozen_stages = 1
+    model = MobileNetv2(frozen_stages=frozen_stages)
+    model.init_weights()
+    model.train()
+    assert model.bn1.training is False
+    for layer in [model.conv1, model.bn1]:
+        for param in layer.parameters():
+            assert param.requires_grad is False
+    for i in range(1, frozen_stages + 1):
+        layer = getattr(model, f'layer{i}')
+        for mod in layer.modules():
+            if isinstance(mod, _BatchNorm):
+                assert mod.training is False
+        for param in layer.parameters():
+            assert param.requires_grad is False
+
+    # Test MobileNetv2 with bn frozen
+    model = MobileNetv2(bn_frozen=True)
+    model.init_weights()
+    model.train()
+    assert model.bn1.training is False
+
+    for i in range(1, 8):
+        layer = getattr(model, f'layer{i}')
+
+        for mod in layer.modules():
+            if isinstance(mod, _BatchNorm):
+                assert mod.training is False
+                for params in mod.parameters():
+                    params.requires_grad = False
+
+    # Test MobileNetv2 forward with widen_factor=1.0
     model = MobileNetv2(widen_factor=1.0, activation=nn.ReLU6)
     model.init_weights()
     model.train()
@@ -20,3 +127,70 @@ def test_mobilenetv2_backbone():
     assert feat[4].shape == torch.Size([1, 96, 14, 14])
     assert feat[5].shape == torch.Size([1, 160, 7, 7])
     assert feat[6].shape == torch.Size([1, 320, 7, 7])
+
+    # Test MobileNetv2 forward with activation=nn.ReLU
+    model = MobileNetv2(widen_factor=1.0, activation=nn.ReLU)
+    model.init_weights()
+    model.train()
+
+    imgs = torch.randn(1, 3, 224, 224)
+    feat = model(imgs)
+    assert len(feat) == 8
+    assert feat[0].shape == torch.Size([1, 16, 112, 112])
+    assert feat[1].shape == torch.Size([1, 24, 56, 56])
+    assert feat[2].shape == torch.Size([1, 32, 28, 28])
+    assert feat[3].shape == torch.Size([1, 64, 14, 14])
+    assert feat[4].shape == torch.Size([1, 96, 14, 14])
+    assert feat[5].shape == torch.Size([1, 160, 7, 7])
+    assert feat[6].shape == torch.Size([1, 320, 7, 7])
+
+    # Test MobileNetv2 with BatchNorm forward
+    model = MobileNetv2(widen_factor=1.0, activation=nn.ReLU6)
+    for m in model.modules():
+        if is_norm(m):
+            assert isinstance(m, _BatchNorm)
+    model.init_weights()
+    model.train()
+
+    imgs = torch.randn(1, 3, 224, 224)
+    feat = model(imgs)
+    assert len(feat) == 8
+    assert feat[0].shape == torch.Size([1, 16, 112, 112])
+    assert feat[1].shape == torch.Size([1, 24, 56, 56])
+    assert feat[2].shape == torch.Size([1, 32, 28, 28])
+    assert feat[3].shape == torch.Size([1, 64, 14, 14])
+    assert feat[4].shape == torch.Size([1, 96, 14, 14])
+    assert feat[5].shape == torch.Size([1, 160, 7, 7])
+    assert feat[6].shape == torch.Size([1, 320, 7, 7])
+
+    # Test MobileNetv2 with layers 1, 3, 5 out forward
+    model = MobileNetv2(
+        widen_factor=1.0, activation=nn.ReLU6, out_indices=(0, 2, 4))
+    model.init_weights()
+    model.train()
+
+    imgs = torch.randn(1, 3, 224, 224)
+    feat = model(imgs)
+    assert len(feat) == 4
+    assert feat[0].shape == torch.Size([1, 16, 112, 112])
+    assert feat[1].shape == torch.Size([1, 32, 28, 28])
+    assert feat[2].shape == torch.Size([1, 96, 14, 14])
+
+    # Test MobileNetv2 with checkpoint forward
+    model = MobileNetv2(widen_factor=1.0, activation=nn.ReLU6, with_cp=True)
+    for m in model.modules():
+        if is_block(m):
+            assert m.with_cp
+    model.init_weights()
+    model.train()
+
+    imgs = torch.randn(1, 3, 224, 224)
+    feat = model(imgs)
+    assert len(feat) == 8
+    assert feat[0].shape == torch.Size([1, 16, 112, 112])
+    assert feat[1].shape == torch.Size([1, 24, 56, 56])
+    assert feat[2].shape == torch.Size([1, 32, 28, 28])
+    assert feat[3].shape == torch.Size([1, 64, 14, 14])
+    assert feat[4].shape == torch.Size([1, 96, 14, 14])
+    assert feat[5].shape == torch.Size([1, 160, 7, 7])
+    assert feat[6].shape == torch.Size([1, 320, 7, 7])