Newest papers

Drive2Gauss: FeedForward 4D Gaussian Reconstruction from Multimodal Generative Driving Videos

Tong Zhao, Zhiyuan Han, Lin Chen, Bowen Xie, Yunxi Qiao, Bohan Li, Weiqing Xiao, Gen Li, Shu Han, Xuyang Dai, Cheng Bi, Hao Zhao, Chaojian Li · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26585 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Menghasilkan lingkungan pengemudi yang dapat digunakan ulang membutuhkan pengamatan masa depan yang dapat dikendalikan dan representasi eksplisit yang mendukung rendering dari sudut pandang yang berbeda, namun metode yang ada masih memulihkan adegan eksplisit dari video RGB yang dihasilkan melalui optimasi untuk setiap urutan. Untuk mengatasi keterbatasan ini, kami memperkenalkan Drive2Gauss, framework feedforward yang mengkonversi pertanyaan RGB yang dihasilkan bersama, kedalaman metrik, dan aliran dinamis menjadi adegan Gaussian 4D yang dinamis. A Multimodal Video Generator memprediksi tiga modalitas dari gambar referensi multiview dan sinyal kontrol masa depan. Selanjutnya, kami memperkenalkan Gaussian Query Decoder yang menggunakan depthalize untuk memulai pertanyaan 3D yang terpecah dan menghubungkan pertanyaan di seluruh waktu, dan mengintegrasikan informasi penampilan melalui agregasi fitur spatiotemp

Machine translation (Meta M2M100); the English original follows.

Abstract

Generating reusable driving environments requires both controllable future observations and an explicit representation that supports rendering from different viewpoints, yet existing methods still recover explicit scenes from generated RGB videos through optimization for each sequence. To overcome these limitations, we present Drive2Gauss, a feedforward framework that converts jointly generated RGB, metric depth, and dynamic flow into dynamic 4D Gaussian scenes. A Multimodal Video Generator predicts the three modalities from reference multiview images and future control signals. Furthermore, we introduce a Gaussian Query Decoder that uses depth to initialize sparse 3D queries and flow to associate the queries across time, and integrates appearance information through spatiotemporal feature aggregation and iterative query refinement. The reconstruction stage operates in a single feedforward pass without optimization for each scene. Extensive results on nuScenes demonstrate that our approach attains an FID of 6.38 and an FVD of 12.43 for future RGB generation, a PSNR of 29.91 dB when reconstructing its generated observations, and supports both direct rendering under lateral camera shifts and downstream video fixing. Code and more visualizations are available at https://zttt04. github.io/Drive2Gauss_page/.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Artificial intelligence, Rendering (computer graphics), Computer vision, Feed forward, RGB color model

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange