Hugging Face Daily Papers · June 30, 2026 · 4 min read

Monte Carlo Energy Aggregation for Mobile 3D Gaussian Splatting

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Like Read original ↗

Code is available: <a href=\"https://github.com/xiaobiaodu/Flux-GS\" rel=\"nofollow\">https://github.com/xiaobiaodu/Flux-GS</a><br>Project page: <a href=\"https://xiaobiaodu.github.io/flux-gs-project/\" rel=\"nofollow\">https://xiaobiaodu.github.io/flux-gs-project/</a></p>\n<p>You can use the idea of our Flux-GS for commercial use for free. We hope it can foster Gaussian Splatting industry development. </p>\n","updatedAt":"2026-06-30T03:12:52.257Z","author":{"_id":"64ad086c5d48838462e2eee1","avatarUrl":"/avatars/735b29d01bae05599e7d5ce61b223153.svg","fullname":"Du","name":"xiaobiaodu","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":1,"identifiedLanguage":{"language":"en","probability":0.8782902359962463},"editors":["xiaobiaodu"],"editorAvatarUrls":["/avatars/735b29d01bae05599e7d5ce61b223153.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.30017","authors":[{"_id":"6a433051763f63ca3757e8e4","name":"Xiaobiao Du","hidden":false},{"_id":"6a433051763f63ca3757e8e5","name":"YuAn Wang","hidden":false},{"_id":"6a433051763f63ca3757e8e6","name":"Hao Li","hidden":false},{"_id":"6a433051763f63ca3757e8e7","name":"Bosheng Wang","hidden":false},{"_id":"6a433051763f63ca3757e8e8","name":"Xun Sun","hidden":false},{"_id":"6a433051763f63ca3757e8e9","name":"Xin Yu","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/64ad086c5d48838462e2eee1/-cKRvS0qOI6mqsYkIjIl5.mp4"],"publishedAt":"2026-06-29T00:00:00.000Z","submittedOnDailyAt":"2026-06-30T00:00:00.000Z","title":"Monte Carlo Energy Aggregation for Mobile 3D Gaussian Splatting","submittedOnDailyBy":{"_id":"64ad086c5d48838462e2eee1","avatarUrl":"/avatars/735b29d01bae05599e7d5ce61b223153.svg","isPro":false,"fullname":"Du","user":"xiaobiaodu","type":"user","name":"xiaobiaodu"},"summary":"Recent advances in 3D Gaussian Splatting have demonstrated unprecedented success in novel view synthesis. However, the substantial inference and storage overhead driven by high-order Spherical Harmonics (SH) are primary bottlenecks for mobile platforms. In this paper, we present Flux-GS, a real-time Gaussian Splatting method designed to achieve high-fidelity rendering with significantly reduced overhead for resource-constrained mobile platforms. We first propose a Monte Carlo Specular Energy Aggregator, sampling third-order radiance residuals and aggregating specular energy into a compact latent space. In this way, our method effectively preserves visually salient lighting features in lower-order bands without expensive distillation or pre-training. To mitigate the high-frequency details lost during compression, we introduce an Attribute-Conditioned SH Enhancement module. This module predicts Gaussian-aware offsets based on intrinsic Gaussian attributes, which enhance the first-order SH representation prior to inference, without extra inference costs. Furthermore, the original single-view gradient-based densification is prone to producing excessive Gaussians and overfitting to a certain view. We address these limitations by proposing a Multi-view Alpha-based Densification and Pruning strategy. By leveraging multi-view guidance, we ensure multi-view structure consistency and the precise removal of redundant primitives. Extensive experiments demonstrate that Flux-GS achieves substantial parameter reduction while maintaining competitive visual quality, offering a robust and scalable solution for real-time mobile rendering. Code: magenta{https://xiaobiaodu.github.io/flux-gs-project/{https://xiaobiaodu.github.io/flux-gs-project/}}.","upvotes":16,"discussionId":"6a433052763f63ca3757e8ea","projectPage":"https://xiaobiaodu.github.io/flux-gs-project/","githubRepo":"https://github.com/xiaobiaodu/Flux-GS","githubRepoAddedBy":"user","ai_summary":"Flux-GS enables real-time high-fidelity 3D Gaussian Splatting on mobile platforms through efficient lighting representation, attribute-conditioned enhancement, and multi-view densification strategies.","ai_keywords":["Spherical Harmonics","Monte Carlo Specular Energy Aggregator","latent space","Attribute-Conditioned SH Enhancement","multi-view gradient-based densification","pruning","Gaussian Splatting","real-time rendering","mobile platforms"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":10,"organization":{"_id":"668de2812413acbd54612298","name":"uts-research","fullname":"University of Technology Sydney","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac1791cb62622e7477d5e2/y07AQFTQdO7Z3fjFFCzj4.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64ad086c5d48838462e2eee1","avatarUrl":"/avatars/735b29d01bae05599e7d5ce61b223153.svg","isPro":false,"fullname":"Du","user":"xiaobiaodu","type":"user"},{"_id":"6985d0c8e650bf6e3993e1bb","avatarUrl":"/avatars/6c0e3ca5be62d8a13998089ca4a19ed1.svg","isPro":false,"fullname":"anonymous","user":"mobile-gs2","type":"user"},{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user"},{"_id":"6a433ac7477f99215df37ed9","avatarUrl":"/avatars/5c0bac9652c66d026222ea95187297da.svg","isPro":false,"fullname":"xixi","user":"huahua11241","type":"user"},{"_id":"673ac9dd2bf4163ef1dc81cf","avatarUrl":"/avatars/de24f8b5c21d4dfa73a0b97e5e395a90.svg","isPro":false,"fullname":"Tianchen Guo","user":"Tianchen5","type":"user"},{"_id":"67641fdd2e02169eb800c2f7","avatarUrl":"/avatars/40143ac788621f7fe405a157d683e565.svg","isPro":false,"fullname":"Shen","user":"Snowmeli","type":"user"},{"_id":"680f8071bab4ac9e563b545f","avatarUrl":"/avatars/53bceeaface4797f19db426596b9d4b3.svg","isPro":false,"fullname":"wangyuan","user":"biubiubiu419","type":"user"},{"_id":"6a4353d51cc8864606dbefbd","avatarUrl":"/avatars/1c87dc88a5880e91efc224266fa9bde9.svg","isPro":false,"fullname":"Hao Li","user":"haoli-china","type":"user"},{"_id":"699c1e69df8352ba379b1172","avatarUrl":"/avatars/faf7fae6a8affc33c2a00246693f1532.svg","isPro":false,"fullname":"cao","user":"uqRu","type":"user"},{"_id":"6407e5294edf9f5c4fd32228","avatarUrl":"/avatars/8e2d55460e9fe9c426eb552baf4b2cb0.svg","isPro":false,"fullname":"Stoney Kang","user":"sikang99","type":"user"},{"_id":"69785dfdad94585f418e9421","avatarUrl":"/avatars/ce4d617b0f45026710a5232e5334f8cf.svg","isPro":false,"fullname":"www","user":"wxz1997","type":"user"},{"_id":"697862fd67aaca70b2c2daaa","avatarUrl":"/avatars/ecb3a79160a6623ef04907c69d7efa18.svg","isPro":false,"fullname":"yizhang","user":"ModelWeaver","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"668de2812413acbd54612298","name":"uts-research","fullname":"University of Technology Sydney","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac1791cb62622e7477d5e2/y07AQFTQdO7Z3fjFFCzj4.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.30017.md","query":{}}">

Papers

arxiv:2606.30017

Monte Carlo Energy Aggregation for Mobile 3D Gaussian Splatting

Published on Jun 29

· Submitted by

Du on Jun 30

University of Technology Sydney

Upvote

Authors:

Abstract

Flux-GS enables real-time high-fidelity 3D Gaussian Splatting on mobile platforms through efficient lighting representation, attribute-conditioned enhancement, and multi-view densification strategies.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

Recent advances in 3D Gaussian Splatting have demonstrated unprecedented success in novel view synthesis. However, the substantial inference and storage overhead driven by high-order Spherical Harmonics (SH) are primary bottlenecks for mobile platforms. In this paper, we present Flux-GS, a real-time Gaussian Splatting method designed to achieve high-fidelity rendering with significantly reduced overhead for resource-constrained mobile platforms. We first propose a Monte Carlo Specular Energy Aggregator, sampling third-order radiance residuals and aggregating specular energy into a compact latent space. In this way, our method effectively preserves visually salient lighting features in lower-order bands without expensive distillation or pre-training. To mitigate the high-frequency details lost during compression, we introduce an Attribute-Conditioned SH Enhancement module. This module predicts Gaussian-aware offsets based on intrinsic Gaussian attributes, which enhance the first-order SH representation prior to inference, without extra inference costs. Furthermore, the original single-view gradient-based densification is prone to producing excessive Gaussians and overfitting to a certain view. We address these limitations by proposing a Multi-view Alpha-based Densification and Pruning strategy. By leveraging multi-view guidance, we ensure multi-view structure consistency and the precise removal of redundant primitives. Extensive experiments demonstrate that Flux-GS achieves substantial parameter reduction while maintaining competitive visual quality, offering a robust and scalable solution for real-time mobile rendering. Code: magenta{https://xiaobiaodu.github.io/flux-gs-project/{https://xiaobiaodu.github.io/flux-gs-project/}}.