Description
This issue proposes adding/documenting support for SHARP – Sharp Monocular View Synthesis in Less Than a Second, a state-of-the-art method for photorealistic novel view synthesis from a single image.

SHARP regresses a metric 3D Gaussian Splatting (3DGS) representation from a single photograph using a single feedforward neural network pass, enabling real-time rendering of nearby views. The approach achieves strong zero-shot generalization across datasets and significantly outperforms prior methods in both quality and inference speed.

Key Highlights

• Single-image → 3D Gaussian Scene in < 1 second
• Metric reconstruction with absolute scale
• Real-time rendering of novel views
• Zero-shot generalization across datasets
• Reduces:

1. LPIPS by 25–34%
2. DISTS by 21–43%

• ~1000× faster synthesis compared to prior work

Proposed Actions

• Add SHARP to supported / referenced view synthesis methods
• Include setup and usage instructions
• Document coordinate conventions and scaling notes