Update transforms.bbox3droi

This method returns a tensor. There was a condition that led to returns of tensors with 7 or 8 size on dim 1 depending on the fulfillment of the condition. The change fixes the shape to 8 without any other changes. This is necessary because the downstream modules that use this output expect a size of 8 on dim 1 not 7. See `Single3DRoIPointExtractor.forward`

mmdet3d/structures/ops/transforms.py

@@ -37,11 +37,12 @@ def bbox3d2roi(bbox_list):
     """
     rois_list = []
     for img_id, bboxes in enumerate(bbox_list):
+        img_inds = bboxes.new_full((bboxes.size(0), 1), img_id)
         if bboxes.size(0) > 0:
-            img_inds = bboxes.new_full((bboxes.size(0), 1), img_id)
             rois = torch.cat([img_inds, bboxes], dim=-1)
         else:
             rois = torch.zeros_like(bboxes)
+            rois = torch.cat([img_inds, rois], dim=-1)
         rois_list.append(rois)
     rois = torch.cat(rois_list, 0)
     return rois