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..
RANK++LETR: Learn to Rank and Optimize Candidates for Line Segment Detection
Authors: Xin Tong, Baojie Tian, Yufei Guo, Zhe Ma
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results demonstrate that our method outperforms previous SOTA methods in prediction accuracy and gets faster inferring speed than other Transformer-based methods. |
| Researcher Affiliation | Collaboration | Xin Tong Baojie Tian Yufei Guo Zhe Ma Intelligent Science & Technology Academy of CASIC EMAIL, EMAIL |
| Pseudocode | No | The paper describes the methodology in natural language and illustrates the overall architecture in Figure 1, but does not provide structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code are provided in the supplemental material. |
| Open Datasets | Yes | We conduct our experiments in two publicly available datasets including the Wireframe dataset [13] and the York Urban dataset [6], which are widely used as line segment detection benchmarks. |
| Dataset Splits | Yes | The Wireframe dataset contains 5,000 training and 462 testing images of man-made environments, while the York Urban dataset contains 102 testing images. The model is only trained on the Wireframe dataset and tested on both Wireframe and York Urban datasets as a typical protocol [14; 45]. |
| Hardware Specification | Yes | We use 4 NVIDIA V100 GPUs for training and 1 GPU for evaluation. |
| Software Dependencies | No | Our training and evaluation are implemented in PyTorch. |
| Experiment Setup | Yes | We train our model for 240 epochs for warming up and 120 epochs for jointly optimizing. The learning rate is set as 5 10 4. The image size and the batch size are set as 512 512 and 8, respectively. We use the Adam W optimizer and set weight decay as 10 4. ... λc, λep, λdp, λr, λs are set to 1, 10, 10, 1, 1, respectively. Moreover, we use auxiliary loss on the early layer in the Transformer-based encoder with a factor of 0.8. K is set to 4 for sampling. |