DäRF: Boosting Radiance Fields from Sparse Input Views with Monocular Depth Adaptation

Authors: Jiuhn Song, Seonghoon Park, Honggyu An, Seokju Cho, Min-Seop Kwak, Sungjin Cho, Seungryong Kim

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show our framework achieves state-of-the-art results both quantitatively and qualitatively, demonstrating consistent and reliable performance in both indoor and outdoor real-world datasets. We evaluate and compare our approach on real-world indoor and outdoor scene datasets, establishing new state-of-the-art results for the benchmarks.
Researcher Affiliation Academia Korea University
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes The code and pre-trained weights will be made publicly available.
Open Datasets Yes Following previous works [37, 45], we use a subset of sparse-view Scan Net data [10] comprised with three indoor scenes... For outdoor reconstruction, we further test on 5 challenging scenes from the Tanks and Temples dataset [19].
Dataset Splits No The paper mentions "training images" and "test images" (e.g., "18 to 20 training images and 8 test images") but does not explicitly provide details about a validation dataset split or how it was used for reproduction.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions several software components like "K-planes [32] as Ne RF", "DPT-hybrid [34] as MDE model", and "Adam [18] as an optimizer," but it does not specify their version numbers or other ancillary software details needed for reproduction.
Experiment Setup Yes We use Adam [18] as an optimizer, with a learning rate of 1 10 2 for Ne RF and 1 10 5 for the MDE, along with a cosine warmup learning rate scheduling.