This report presents the approach used in the submission of the LVIS Challenge 2020 of team MMDet. In the submission, we propose seesaw loss that dynamically rebalances the penalty to each category according to a relative ratio of cumulative training instances between different categor