SESS: Self-Ensembling Semi-Supervised 3D Object Detection

In this appendix, we provide performance comparison between SESS and VoteNet with more diverse ratios of labeled data on the SUN RGB-D and ScaNetV2 val sets in Sec. A. We also provide additional evaluation metric (i.e. mAP@0.5 IoU) for both inductive and transductive semisupervised learning in Sec. B. In Sec. C, we report per-class average precision on the SUN RGB-D and ScanNetV2 val set. Finally, more qualitative results are shown in Section D.