Problem: Currently, SHARP produces high-quality 3D reconstructions from single images, but the output is primarily consumed within the provided inference scripts. There is no straightforward way to use these results in external 3D software.
Proposed Solution: Add a utility script or an argument in predict.py to export the generated 3D Gaussian Splats into industry-standard formats such as:
. ply: For compatibility with Gaussian Splatting viewers (like Polycam or SuperSplat).
. splat: For web-based WebGL renderers.
Value: This would enable researchers and creators to integrate SHARP into their 3D pipelines (Blender, Unity, Unreal Engine) immediately after inference.