Add CecaBam

CecaBam is circular ECA followed by BAM for both channelwise and spatial attention

timm/models/EcaModule.py

@@ -32,6 +32,7 @@ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 SOFTWARE.
 '''
 import math
+import torch
 from torch import nn
 import torch.nn.functional as F
 
@@ -119,3 +120,33 @@ class CecaModule(nn.Module):
         y = self.sigmoid(y.view(x.shape[0], -1, 1, 1))
 
         return x * y.expand_as(x)
+
+class ChannelPool(nn.Module):
+    def forward(self, x):
+        return torch.cat((torch.max(x,1)[0].unsqueeze(1), torch.mean(x,1).unsqueeze(1)), dim=1)
+
+class SpatialGate(nn.Module):
+    def __init__(self):
+        super(SpatialGate, self).__init__()
+        kernel_size = 7
+        self.compress = ChannelPool()
+        self.spatial = nn.Conv2d(2, 1, kernel_size, stride=1, padding=(kernel_size-1) // 2, bias=False)
+    def forward(self, x):
+        x_compress = self.compress(x)
+        x_out = self.spatial(x_compress)
+        scale = torch.sigmoid(x_out) # broadcasting
+        return x * scale
+
+
+class CecaBam(nn.Module):
+    def __init__(self, gate_channels, no_spatial=False):
+            super(CecaBam, self).__init__()
+            self.CecaModule = CecaModule(gate_channels)
+            self.no_spatial=no_spatial
+            if not no_spatial:
+                self.SpatialGate = SpatialGate()
+    def forward(self, x):
+        x_out = self.CecaModule(x)
+        if not self.no_spatial:
+            x_out = self.SpatialGate(x_out)
+        return x_out