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..
RankSEG-RMA: An Efficient Segmentation Algorithm via Reciprocal Moment Approximation
Authors: Zixun Wang, Ben Dai
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We illustrate the effectiveness of our method across various datasets and state-of-the-art models. The code of our method is available in: https://github.com/Zixun Wang/Rank SEG-RMA. 4 Experiments Datasets. We conduct experiments on five datasets: (1) PASCAL VOC [Everingham et al., 2010], (2) Cityscapes [Cordts et al., 2016], (3) ADE20K [Zhou et al., 2017], (4) Li TS [Bilic et al., 2023], and (5) Ki TS [Heller et al., 2021]. |
| Researcher Affiliation | Academia | Zixun Wang Department of Statistics and Data Science The Chinese University of Hong Kong EMAIL Ben Dai Department of Statistics and Data Science The Chinese University of Hong Kong EMAIL |
| Pseudocode | Yes | Algorithm 1 Rank Dice-RMA-Binary Input: Estimated probability map bp [0, 1]d for a given input image. Output: The predicted segmentation mask bĪ“ {0, 1}d. |
| Open Source Code | Yes | The code of our method is available in: https://github.com/Zixun Wang/Rank SEG-RMA. |
| Open Datasets | Yes | Datasets. We conduct experiments on five datasets: (1) PASCAL VOC [Everingham et al., 2010], (2) Cityscapes [Cordts et al., 2016], (3) ADE20K [Zhou et al., 2017], (4) Li TS [Bilic et al., 2023], and (5) Ki TS [Heller et al., 2021]. |
| Dataset Splits | Yes | Furthermore, since Li TS and Ki TS do not include designated test sets, we employ 5-fold cross-validation to evaluate performance, following existing literature [Qin et al., 2021]. |
| Hardware Specification | Yes | Table 3: Time consumption (in seconds) of model forward and different prediction rules with single A100 GPU. |
| Software Dependencies | No | The paper mentions optimizers like "Adam W optimizer" and "SGD", and general learning rate policies, but does not provide specific version numbers for any libraries or frameworks (e.g., PyTorch, TensorFlow, scikit-learn). The training details are provided, but without specific version numbers for the software used, it's not considered a reproducible description of ancillary software. |
| Experiment Setup | Yes | The training settings mainly follow Wang et al. [2023a,b]. For Pascal VOC, Cityscapes and ADE20K, Adam W optimizer with a weight decay of 0.01 is used. The learning rate starts from 1e 6 and linearly warms up during the first 1% iterations to the initial learning rate 6e 5. The learning rate is then decayed in a poly policy with an exponent of 1. The number of warm-up iterations is 400 for Pascal VOC and Cityscapes, and 800 for ADE20K. The total number of training iterations is 40,000 for Pascal VOC and Cityscapes, and 80,000 for ADE20K. Data augmentations including (i) random scaling in the range of [0.5, 2.0], and (ii) random horizontal flipping with a probability of 0.5. For Li TS and Ki TS, we train the models using SGD with an initial learning rate of 0.01, momentum of 0.9, and weight decay of 0.0005. The learning rate is decayed in a poly policy with an exponent of 0.9. The batch size is 8 and the number of epochs is 60. |