Fix dpool (#1390)

* fix dpool

* add _pair in dpool func

mmdet/ops/dcn/deform_pool.py

@@ -2,6 +2,7 @@ import torch
 import torch.nn as nn
 from torch.autograd import Function
 from torch.autograd.function import once_differentiable
+from torch.nn.modules.utils import _pair
 
 from . import deform_pool_cuda
 
@@ -21,6 +22,12 @@ class DeformRoIPoolingFunction(Function):
                 part_size=None,
                 sample_per_part=4,
                 trans_std=.0):
+        # TODO: support unsquare RoIs
+        out_h, out_w = _pair(out_size)
+        assert isinstance(out_h, int) and isinstance(out_w, int)
+        assert out_h == out_w
+        out_size = out_h  # out_h and out_w must be equal
+
         ctx.spatial_scale = spatial_scale
         ctx.out_size = out_size
         ctx.out_channels = out_channels
@@ -85,7 +92,7 @@ class DeformRoIPooling(nn.Module):
                  trans_std=.0):
         super(DeformRoIPooling, self).__init__()
         self.spatial_scale = spatial_scale
-        self.out_size = out_size
+        self.out_size = _pair(out_size)
         self.out_channels = out_channels
         self.no_trans = no_trans
         self.group_size = group_size
@@ -125,12 +132,12 @@ class DeformRoIPoolingPack(DeformRoIPooling):
 
         if not no_trans:
             seq = []
-            ic = self.out_size * self.out_size * self.out_channels
+            ic = self.out_size[0] * self.out_size[1] * self.out_channels
             for i in range(self.num_offset_fcs):
                 if i < self.num_offset_fcs - 1:
                     oc = self.deform_fc_channels
                 else:
-                    oc = self.out_size * self.out_size * 2
+                    oc = self.out_size[0] * self.out_size[1] * 2
                 seq.append(nn.Linear(ic, oc))
                 ic = oc
                 if i < self.num_offset_fcs - 1:
@@ -156,7 +163,7 @@ class DeformRoIPoolingPack(DeformRoIPooling):
                                    self.group_size, self.part_size,
                                    self.sample_per_part, self.trans_std)
             offset = self.offset_fc(x.view(n, -1))
-            offset = offset.view(n, 2, self.out_size, self.out_size)
+            offset = offset.view(n, 2, self.out_size[0], self.out_size[1])
             return deform_roi_pooling(data, rois, offset, self.spatial_scale,
                                       self.out_size, self.out_channels,
                                       self.no_trans, self.group_size,
@@ -188,12 +195,12 @@ class ModulatedDeformRoIPoolingPack(DeformRoIPooling):
 
         if not no_trans:
             offset_fc_seq = []
-            ic = self.out_size * self.out_size * self.out_channels
+            ic = self.out_size[0] * self.out_size[1] * self.out_channels
             for i in range(self.num_offset_fcs):
                 if i < self.num_offset_fcs - 1:
                     oc = self.deform_fc_channels
                 else:
-                    oc = self.out_size * self.out_size * 2
+                    oc = self.out_size[0] * self.out_size[1] * 2
                 offset_fc_seq.append(nn.Linear(ic, oc))
                 ic = oc
                 if i < self.num_offset_fcs - 1:
@@ -203,12 +210,12 @@ class ModulatedDeformRoIPoolingPack(DeformRoIPooling):
             self.offset_fc[-1].bias.data.zero_()
 
             mask_fc_seq = []
-            ic = self.out_size * self.out_size * self.out_channels
+            ic = self.out_size[0] * self.out_size[1] * self.out_channels
             for i in range(self.num_mask_fcs):
                 if i < self.num_mask_fcs - 1:
                     oc = self.deform_fc_channels
                 else:
-                    oc = self.out_size * self.out_size
+                    oc = self.out_size[0] * self.out_size[1]
                 mask_fc_seq.append(nn.Linear(ic, oc))
                 ic = oc
                 if i < self.num_mask_fcs - 1:
@@ -236,9 +243,9 @@ class ModulatedDeformRoIPoolingPack(DeformRoIPooling):
                                    self.group_size, self.part_size,
                                    self.sample_per_part, self.trans_std)
             offset = self.offset_fc(x.view(n, -1))
-            offset = offset.view(n, 2, self.out_size, self.out_size)
+            offset = offset.view(n, 2, self.out_size[0], self.out_size[1])
             mask = self.mask_fc(x.view(n, -1))
-            mask = mask.view(n, 1, self.out_size, self.out_size)
+            mask = mask.view(n, 1, self.out_size[0], self.out_size[1])
             return deform_roi_pooling(
                 data, rois, offset, self.spatial_scale, self.out_size,
                 self.out_channels, self.no_trans, self.group_size,