Would you please be kind enough to share your quantitative results for the ResNet18 backbone Depth model with velocity supervision? Thank you