AnySplat: Feed-forward 3D Gaussian Splatting from Unconstrained Views

TL;DR

AnySplat is a feed-forward neural network that synthesizes novel views from uncalibrated images by predicting 3D Gaussian primitives in a single pass, enabling real-time rendering without pose annotations.

Contribution

It introduces a unified, pose-free neural rendering approach that scales to casual multi-view datasets and matches pose-aware methods in quality.

Findings

Achieves comparable quality to pose-aware methods in zero-shot evaluations.

Reduces rendering latency significantly, enabling real-time synthesis.

Operates effectively on uncalibrated, casually captured multi-view datasets.

Abstract

We introduce AnySplat, a feed forward network for novel view synthesis from uncalibrated image collections. In contrast to traditional neural rendering pipelines that demand known camera poses and per scene optimization, or recent feed forward methods that buckle under the computational weight of dense views, our model predicts everything in one shot. A single forward pass yields a set of 3D Gaussian primitives encoding both scene geometry and appearance, and the corresponding camera intrinsics and extrinsics for each input image. This unified design scales effortlessly to casually captured, multi view datasets without any pose annotations. In extensive zero shot evaluations, AnySplat matches the quality of pose aware baselines in both sparse and dense view scenarios while surpassing existing pose free approaches. Moreover, it greatly reduce rendering latency compared to optimization based neural fields, bringing real time novel view synthesis within reach for unconstrained capture settings.Project page: https://city-super.github.io/anysplat/