Overflow Error in Coarse Rasterization Phase When Converting Gsplat to Mesh with SugaR

[jj7258]

Description First, thank you for creating such a wonderful project!

I’ve been trying to use [SugaR] to convert Gsplat models into mesh files. During the mesh generation training pipeline, I encountered an overflow error in the coarse rasterization phase. The command I ran was:

python train.py -s <COLMAP dataset path> -r "dn_consistency" --high_poly True --refinement_time "long" -c <GS Checkpoint Path>

---

Observed Behavior The console prints:

Bin size was too small in the coarse rasterization phase. This caused an overflow, meaning output may be incomplete.
To solve, try increasing max_faces_per_bin / max_points_per_bin, decreasing bin_size, or setting bin_size to 0 to use the naive rasterization.

---

Request

1. Has anyone encountered and resolved this overflow error?
2. Are there alternative rasterization strategies or flags that work better for high‐density Gsplat models?
3. Is it possible to import Gsplat files generated by other pipelines (e.g., gaussian-splatting-lightning) into the SugaR training/conversion pipeline?

Thank you for any pointers! 😊