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

2026-03-11

AI Summary

This work proposes InstantHDR, the first feed-forward neural network capable of reconstructing high-dynamic-range (HDR) 3D scenes from uncalibrated multi-exposure low-dynamic-range (LDR) images in a single forward pass, eliminating the need for per-scene optimization. Existing HDR reconstruction methods typically rely on known camera poses, dense point cloud initialization, and time-consuming optimization. In contrast, InstantHDR introduces a geometry-guided multi-exposure fusion mechanism, a generalizable meta-learned tone-mapping network, and an efficient Gaussian splatting-based scene representation. To support training, the authors also present HDR-Pretrain, the first large-scale synthetic dataset tailored for feed-forward HDR reconstruction. Experiments demonstrate that InstantHDR achieves reconstruction quality comparable to state-of-the-art optimization-based methods while offering speedups of approximately 700× without post-optimization or 20× with lightweight refinement.

Technology Category

• Computer Vision: 3D Computer Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

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.

Problem

• HDR
• novel view synthesis
• multi-exposure
• 3D reconstruction
• feed-forward

Innovation

• feed-forward HDR reconstruction
• geometry-guided appearance modeling
• meta-network tone mapping
• multi-exposure fusion
• HDR-Pretrain dataset