InstantHDR: Single-forward Gaussian Splatting for High Dynamic Range 3D Reconstruction

TL;DR

InstantHDR is a fast, feed-forward method for high dynamic range 3D scene reconstruction from multi-exposure images, achieving comparable quality to traditional methods with significantly improved speed.

Contribution

The paper introduces InstantHDR, a novel single-forward pass neural network for HDR 3D reconstruction that does not require scene-specific optimization or known camera poses.

Findings

Achieves similar quality to state-of-the-art optimization-based HDR methods.

Provides approximately 700x faster reconstruction with single-forward pass.

Uses a new HDR-Pretrain dataset for generalizable training.

Abstract

High dynamic range (HDR) novel view synthesis (NVS) aims to reconstruct HDR scenes from multi-exposure low dynamic range (LDR) images. Existing HDR pipelines heavily rely on known camera poses, well-initialized dense point clouds, and time-consuming per-scene optimization. Current feed-forward alternatives overlook the HDR problem by assuming exposure-invariant appearance. To bridge this gap, we propose InstantHDR, a feed-forward network that reconstructs 3D HDR scenes from uncalibrated multi-exposure LDR collections in a single forward pass. Specifically, we design a geometry-guided appearance modeling for multi-exposure fusion, and a meta-network for generalizable scene-specific tone mapping. Due to the lack of HDR scene data, we build a pre-training dataset, called HDR-Pretrain, for generalizable feed-forward HDR models, featuring 168 Blender-rendered scenes, diverse lighting types, and multiple camera response functions. Comprehensive experiments show that our InstantHDR delivers comparable synthesis performance to the state-of-the-art optimization-based HDR methods while enjoying $\sim700\times$ and $\sim20\times$ reconstruction speed improvement with our single-forward and post-optimization settings. All code, models, and datasets will be released after the review process.