EcoSplat: Efficiency-controllable Feed-forward 3D Gaussian Splatting from Multi-view Images

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Park, Jongmin, Bui, Minh-Quan Viet, Bello, Juan Luis Gonzalez, Moon, Jaeho, Oh, Jihyong, Kim, Munchurl
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912780399935488
author Park, Jongmin
Bui, Minh-Quan Viet
Bello, Juan Luis Gonzalez
Moon, Jaeho
Oh, Jihyong
Kim, Munchurl
author_facet Park, Jongmin
Bui, Minh-Quan Viet
Bello, Juan Luis Gonzalez
Moon, Jaeho
Oh, Jihyong
Kim, Munchurl
contents Feed-forward 3D Gaussian Splatting (3DGS) enables efficient one-pass scene reconstruction, providing 3D representations for novel view synthesis without per-scene optimization. However, existing methods typically predict pixel-aligned primitives per-view, producing an excessive number of primitives in dense-view settings and offering no explicit control over the number of predicted Gaussians. To address this, we propose EcoSplat, the first efficiency-controllable feed-forward 3DGS framework that adaptively predicts the 3D representation for any given target primitive count at inference time. EcoSplat adopts a two-stage optimization process. The first stage is Pixel-aligned Gaussian Training (PGT) where our model learns initial primitive prediction. The second stage is Importance-aware Gaussian Finetuning (IGF) stage where our model learns rank primitives and adaptively adjust their parameters based on the target primitive count. Extensive experiments across multiple dense-view settings show that EcoSplat is robust and outperforms state-of-the-art methods under strict primitive-count constraints, making it well-suited for flexible downstream rendering tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2512_18692
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle EcoSplat: Efficiency-controllable Feed-forward 3D Gaussian Splatting from Multi-view Images
Park, Jongmin
Bui, Minh-Quan Viet
Bello, Juan Luis Gonzalez
Moon, Jaeho
Oh, Jihyong
Kim, Munchurl
Computer Vision and Pattern Recognition
Feed-forward 3D Gaussian Splatting (3DGS) enables efficient one-pass scene reconstruction, providing 3D representations for novel view synthesis without per-scene optimization. However, existing methods typically predict pixel-aligned primitives per-view, producing an excessive number of primitives in dense-view settings and offering no explicit control over the number of predicted Gaussians. To address this, we propose EcoSplat, the first efficiency-controllable feed-forward 3DGS framework that adaptively predicts the 3D representation for any given target primitive count at inference time. EcoSplat adopts a two-stage optimization process. The first stage is Pixel-aligned Gaussian Training (PGT) where our model learns initial primitive prediction. The second stage is Importance-aware Gaussian Finetuning (IGF) stage where our model learns rank primitives and adaptively adjust their parameters based on the target primitive count. Extensive experiments across multiple dense-view settings show that EcoSplat is robust and outperforms state-of-the-art methods under strict primitive-count constraints, making it well-suited for flexible downstream rendering tasks.
title EcoSplat: Efficiency-controllable Feed-forward 3D Gaussian Splatting from Multi-view Images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.18692