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..
UniMRSeg: Unified Modality-Relax Segmentation via Hierarchical Self-Supervised Compensation
Authors: Xiaoqi Zhao, Youwei Pang, Chenyang Yu, Lihe Zhang, Huchuan Lu, Shijian Lu, Georges Fakhri, Xiaofeng Liu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this work, we conduct experimental comparisons on four popular visual multi-modal image segmentation tasks to show the generalizability of the proposed method. |
| Researcher Affiliation | Academia | 1Yale University, USA 2 Nanyang Technological University, Singapore 3Dalian University of Technology, China |
| Pseudocode | No | The paper describes methods in text and uses diagrams (e.g., Figure 3) to illustrate the framework, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The code will be released at https://github.com/Xiaoqi-Zhao-DLUT/Uni MRSeg. |
| Open Datasets | Yes | I) Brain Tumor Segmentation. We follow most brain tumor segmentation methods [22, 63, 10, 30] use the Bra TS2020 dataset [32]... II) RGB-D Salient Object Segmentation. We adopt the same training set as most methods [80, 82, 39], i.e., 1,485 samples from the NJUD [20] and 700 samples from the NLPR [43]. The test dataset is STERE [34]... III) RGB-T Salient Object Segmentation. We follow the setting of recent works [39, 86, 51], the training set only contains the 2,500 samples from VT5000 [53] and adopt the VT1000 [54] as the test set... IV) RGB-D Semantic Segmentation. SUN-RGBD [50] is the popular indoor scene benchmark... |
| Dataset Splits | Yes | The dataset is split into training (315), validation (17), and test (37) sets. ... It contains 10,335 pairs of RGB-D images, with 5,285 pairs allocated for training and 5,050 for testing. ... for Bra TS2020, where the dataset is divided into 219 cases for training, 50 for validation, and 100 for testing. For Bra TS2018, we follow the data split protocol from M3Fe Con [69], using 200 cases for training and 85 for testing. |
| Hardware Specification | Yes | All experiments are conducted on one NVIDIA A800 GPU. |
| Software Dependencies | No | The paper mentions using the 'Adam W optimizer [31]' but does not provide specific version numbers for any software libraries or dependencies used in the implementation, such as Python, PyTorch, or CUDA. |
| Experiment Setup | Yes | We train the model for 300 epochs based on the Adam W optimizer [31] with a warmup schedule, an initial learning rate of 0.0001, and a weight decay of 0.00001. |