diff --git a/mmseg/models/segmentors/base.py b/mmseg/models/segmentors/base.py
index 25487de5a..17a0bb2b3 100644
--- a/mmseg/models/segmentors/base.py
+++ b/mmseg/models/segmentors/base.py
@@ -187,6 +187,7 @@ class BaseSegmentor(BaseModel, metaclass=ABCMeta):
             if C > 1:
                 i_seg_pred = i_seg_logits.argmax(dim=0, keepdim=True)
             else:
+                i_seg_logits = i_seg_logits.sigmoid()
                 i_seg_pred = (i_seg_logits >
                               self.decode_head.threshold).to(i_seg_logits)
             data_samples[i].set_data({
diff --git a/mmseg/models/segmentors/encoder_decoder.py b/mmseg/models/segmentors/encoder_decoder.py
index 0a8db3ec7..370d0305f 100644
--- a/mmseg/models/segmentors/encoder_decoder.py
+++ b/mmseg/models/segmentors/encoder_decoder.py
@@ -260,10 +260,10 @@ class EncoderDecoder(BaseSegmentor):
         h_stride, w_stride = self.test_cfg.stride
         h_crop, w_crop = self.test_cfg.crop_size
         batch_size, _, h_img, w_img = inputs.size()
-        num_classes = self.num_classes
+        out_channels = self.out_channels
         h_grids = max(h_img - h_crop + h_stride - 1, 0) // h_stride + 1
         w_grids = max(w_img - w_crop + w_stride - 1, 0) // w_stride + 1
-        preds = inputs.new_zeros((batch_size, num_classes, h_img, w_img))
+        preds = inputs.new_zeros((batch_size, out_channels, h_img, w_img))
         count_mat = inputs.new_zeros((batch_size, 1, h_img, w_img))
         for h_idx in range(h_grids):
             for w_idx in range(w_grids):