train YOLOv2 on VOC

models/yolov2/loss.py

@@ -82,7 +82,7 @@ class Criterion(object):
         loss_box = loss_box.sum() / num_fgs
         
         # obj loss
-        gt_objectness[pos_masks] *= ious.clamp(0.)
+        # gt_objectness[pos_masks] *= ious.clamp(0.)
         loss_obj = self.loss_objectness(pred_obj, gt_objectness)
         loss_obj = loss_obj.sum() / num_fgs