BungeeNeRF: Progressive Neural Radiance Field for Extreme Multi-scale Scene Rendering

Xiangli, Yuanbo; Xu, Liying; Pan, Xingang; Zhao, Nanxuan; Rao, Anyi; Theobalt, Christian; Dai, Bo; Lin, Dahua

doi:10.1007/978-3-031-19824-3_7

Cited by 127 publications

(47 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…synthesis performance will decrease drastically if migrated to outdoor scenes. Although some very recent methods try to solve this problem and are well-designed for large scenes [35,39], their performance is difficult to compare due to the lack of large-scale outdoor scene datasets and uniform benchmarks.…”

Section: More Scenesmentioning

confidence: 99%

“…Figure 2. Visual comparison with existing large-scale satelliteview outdoor datasets [39] acquired from Google Earth Studio. The top row is from [39], and the bottom row is corresponding scenes from our fly-view dataset, which is more realistic with clear textures and rich details (zoom-in for the best of views).…”

Section: Satellite-viewmentioning

confidence: 99%

“…Block-NeRF [34] also adopts this simplified idea, dividing the neighborhood into blocks, and then novel views are sampled from overlapping blocks and combined according to inverse distance weights. CityNerf (BungeeNeRF) [39] introduces a progressive neural radiance field that starts from fitting distant views with a shallow base block, and appends new blocks to accommodate details. NeRF in the Wild (NeRF-W) [23] introduces a series of extensions to NeRF [24] to synthesize novel views of complex scenes, using only unstructured collections of in-the-wild photos.…”

Section: Neural Scene Representation and Renderingmentioning

confidence: 99%

“…However, the aforementioned large-scale NeRF methods [35,34,39,23,40]use different outdoor datasets with various capturing conditions and research focuses, causing difficulty to fairly compare these NeRF methods' performance on common tasks based on a uniform benchmark.…”

Section: Neural Scene Representation and Renderingmentioning

confidence: 99%

“…Mega-NeRF [35] builds two high-quality fly-view real scenes and calibrated images, but has not become a widely used benchmark due to its small size and single type. There are also large-scale NeRF datasets and benchmarks for specific problems, such as neighborhoods [34] or remote sensing [39], see the second group in Table 1.…”

Section: Nerf-based Datasets and Benchmarksmentioning

confidence: 99%

See 4 more Smart Citations

A Large-Scale Outdoor Multi-modal Dataset and Benchmark for Novel View Synthesis and Implicit Scene Reconstruction

Lu¹,

Yin²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

Neural Radiance Fields (NeRF) has achieved impressive results in single object scene reconstruction and novel view synthesis, as demonstrated on many single modality and single object focused indoor scene datasets like DTU [14], BMVS [41], and NeRF Synthetic [24]. However, the study of NeRF on large-scale outdoor scene reconstruction is still limited, as there is no unified outdoor scene dataset for large-scale NeRF evaluation due to expensive data acquisition and calibration costs.In this work, we propose a large-scale outdoor multimodal dataset, OMMO dataset, containing complex objects and scenes with calibrated images, point clouds and prompt annotations. A new benchmark for several outdoor NeRF-based tasks is established, such as novel view synthesis, surface reconstruction, and multi-modal NeRF. To create the dataset, we capture and collect a large number of real fly-view videos and select high-quality and highresolution clips from them. Then we design a quality review module to refine images, remove low-quality frames and fail-to-calibrate scenes through a learning-based automatic evaluation plus manual review. Finally, a number of volunteers are employed to add the text descriptions for each scene and keyframe.Compared with existing NeRF datasets, our dataset contains abundant real-world urban and natural scenes with various scales, camera trajectories, and lighting conditions. Experiments show that our dataset can benchmark most state-of-the-art NeRF methods on different tasks. We will release the dataset and model weights soon.

show abstract

Section: More Scenesmentioning

confidence: 99%

Section: Satellite-viewmentioning

confidence: 99%