未验证 提交 c73a5814 编写于 作者: T Tian Zhi 提交者: GitHub

fixed a bug.

上级 4697f876
......@@ -105,8 +105,8 @@ class FCOSHead(torch.nn.Module):
bbox_pred = self.scales[l](self.bbox_pred(box_tower))
if self.norm_reg_targets:
bbox_pred = F.relu(bbox_pred)
if self.training:
bbox_pred = F.relu(bbox_pred)
bbox_reg.append(bbox_pred)
else:
bbox_reg.append(bbox_pred * self.fpn_strides[l])
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册