Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Learnable Burst-Encodable Time-of-Flight Imaging for High-Fidelity Long-Distance Depth Sensing
Authors: Manchao Bao, Shengjiang Fang, Tao Yue, Xuemei Hu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The proposed approach is rigorously validated through comprehensive simulations and real-world prototype experiments, demonstrating its effectiveness and practical applicability. The code is available at: https://github.com/Computational Perception Lab/BE-To F. In this paper, we propose a novel To F imaging paradigm termed Burst-Encodable Time-of-Flight (BE-To F). ... We evaluate our method on a synthetic dataset and compare it with conventional i To F approaches, including single-frequency and multi-frequency modulation techniques. Finally, we built a prototype system to prove the effectiveness of our method in real-world experiments. |
| Researcher Affiliation | Academia | 1Nanjing University 2Nanjing Electronic Devices Institute |
| Pseudocode | No | The paper describes the methodology using mathematical formulations and descriptive text, but it does not include any clearly labeled pseudocode or algorithm blocks. For example, Section 3 'Learnable Burst-Encodable Time-of-Flight Imaging' describes the working principle and models, and Section 3.2 'Reconstruction Network' details the network architecture, but none are presented as formal pseudocode. |
| Open Source Code | Yes | The code is available at: https://github.com/Computational Perception Lab/BE-To F. |
| Open Datasets | Yes | We use the NYU-V2 dataset [37] to train and test our end-to-end framework. The NYU-V2 dataset is a high-quality RGB-D dataset captured by Kinect. It contains a total of 1449 pairs of precisely aligned RGB and depth images collected from 464 indoor scenes, which enables its extensive application in academic research. ... [37] Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and support inference from rgbd images. In Proceedings of the European Conference on Computer Vision, pages 746 760. Springer, 2012. |
| Dataset Splits | Yes | We divide the dataset in detail, using 1000 pairs of data as the training set and the remaining 449 pairs as the test set [39, 40]. |
| Hardware Specification | Yes | All experiments are conducted on the Py Torch platform [43], using an NVIDIA Ge Force RTX 4090 GPU. ... to validate the effectiveness of our BE-To F approach in real world scenarios, we built a prototype system comprising a solid-state pulsed laser and an exposure-encodable ICMOS sensor. The laser operates at 532 nm with a 5 ns pulse width, a fixed 1 k Hz repetition rate, and up to 1 m J single-pulse energy. ... The ICMOS is fitted with a zoom lens (300-800 mm) and supports a minimum exposure gate of 3 ns. |
| Software Dependencies | No | All experiments are conducted on the Py Torch platform [43], using an NVIDIA Ge Force RTX 4090 GPU. The paper mentions PyTorch but does not provide a specific version number. No other software components with version numbers are explicitly listed. |
| Experiment Setup | Yes | We choose K in Eq. 6 as 4 and M in Eq. 11 as 1000. The number of restormer blocks in the network is set to [4, 6, 6, 8]. We train the network for 200 epochs using the ADAM optimizer [42] with a batch size of 20. The learning rate is initialized at 0.01 and decays by a factor of 0.7 every 10 epochs. The loss balance coefficients γ1 and γ2 are empirically set to 5e-4 and 5e-2 initially, and are updated to 5e-5 and 1 after 40 epochs. γ3 is always set to 5. Xavier initialization is used for the learnable coding functions. |