About leveraging sparse LiDAR priors with Depth Anything V3 for metric depth estimation

Nice work bro. I'm a researcher working on metric depth estimation, and I've been very impressed by your recent work on Depth Anything V3. I'm writing to ask about the possibility of using sparse point cloud priors with your model.

**My Setup:**
I'm working with a dataset (ConSLAM) that provides synchronized RGB images and sparse LiDAR point clouds from a 16/32-line scanner. Each frame has:
- RGB image (2058×1533)
- Sparse point cloud (.pcd)
- Accurate camera intrinsics and extrinsics

**What I've Tried:**
I've successfully run inference with DA3-BASE on single images and obtained relative depth maps. However, I noticed that the `inference()` API doesn't accept any form of depth prior (like sparse depth maps).

**My Questions:**
1. Does Depth Anything V3 currently support any way to incorporate external depth priors (e.g., sparse LiDAR points) to guide metric depth estimation?
2.If doesn't, is there any alternative approach you would recommend for my use case (image + sparse LiDAR points → dense metric depth)?

Thank you very much for your time and for sharing your excellent work with the community. Any guidance would be greatly appreciated.