### What's the feature? Thank you for your contribution. Why don't you extract pseudo labels by combining the boxes predicted by the normal, depth, and RGB models? ### Any other context? _No response_