KATTA HAJMLI SAHNALARNI 3D GAUSSIAN SPLATTING ASOSIDA REAL VAQTDA VIZUALLASHTIRISHDA XOTIRA VA HISOBLASH RESURSLARINI OPTIMALLASHTIRISH
Jild 5 son 26 (2026): Zamonaviy dunyoda innovatsion tadqiqotlar 24-27
Annotatsiya
3D Gaussian Splatting (3DGS) real vaqtda yuqori sifatli yangi ko‘rinishlarni sintez qilishda inqilob qilgan bo‘lsa-da, uni katta hajmli sahnalarga (masalan, shahar ko‘chalari, me’moriy majmualar, aerofotosuratlar asosidagi hududlar) qo‘llashda Gauss primitivlari sonining keskin ortishi xotira va hisoblash resurslari muammosini keltirib chiqaradi. Zamonaviy GPU xotirasi (odatda 24GB) katta hajmli sahnalarni yuqori aniqlikda o‘qitish va render qilish uchun yetarli emas. Ushbu tezisda katta hajmli meros obyektlarini real vaqtda vizuallashtirishda resurslarni optimallashtirishning kompleks strategiyasi taklif etiladi. Strategiya uchta asosiy komponentdan iborat: (1) kameraga masofaga asoslangan Level-of-Detail (LOD) usuli – optimal Gauss to‘plamlarini iterativ tanlash orqali render vaqti va GPU xotirasini kamaytiradi; (2) visibility-in-block bo‘linish strategiyasi – sahna bloklarga bo‘linib, faqat ko‘rinadigan piksellar o‘qitishga jalb qilinadi; (3) host offloading mexanizmi – Gauss primitivlari host xotirasida saqlanib, faqat kerakli qismi GPUga yuklanadi.
Kalit soʻzlar:
Iqtiboslar
Kerbl, B., Kopanas, G., Leimkühler, T., & Drettakis, G. (2023). 3D Gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4), 1-14.
Kulhanek, J., Rakotosaona, M. J., Manhardt, F., Tsalicoglou, C., Niemeyer, M., Sattler, T., Peng, S., & Tombari, F. (2025). LODGE: Level-of-detail large-scale Gaussian splatting with efficient rendering. NeurIPS 2025.
Shen, J., Qian, Y., & Zhan, X. (2025). LOD-GS: Achieving levels of detail using scalable Gaussian soup. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 671-680.
GS-Scale (2025). GS-Scale: Unlocking large-scale 3D Gaussian splatting training via host offloading. arXiv preprint
Octree-GS (2025). Octree-GS: Towards consistent real-time rendering with LOD-structured 3D Gaussians. IEEE Transactions on Visualization and Computer Graphics.
CityGS-X (2025). CityGS-X: A scalable architecture for efficient and geometrically accurate large-scale scene reconstruction. ICCV 2025.
Consistency-preserving Gaussian splatting for block-based large-scale scene reconstruction (2025). Elsevier.
Chen, G., & Wang, W. (2026). A survey on 3D Gaussian splatting. ACM Computing Surveys, 58(12), 1-39.
Yuklab olishlar
Nashr qilingan
Son
Boʻlim
Litsenziya
##submission.license.cc.by4.footer##Iqtibos keltirish tartibi
