# UnfoldCRF: Structured Mask Refinement with Image-Conditioned Latent Regions

arXiv: https://arxiv.org/abs/2609.33996 · HTML: https://mlsanity.com/paper/2609.33996/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 32 of its 50 references that are in the map.
- First posted on arXiv 27 Sep 2026.

## Builds on (most cited first, 30 of 32)

- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Fully Convolutional Networks for Semantic Segmentation](https://mlsanity.com/paper/1411.4038/) (2014)
- [Rethinking Atrous Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1706.05587/) (2017)
- [Searching for MobileNetV3](https://mlsanity.com/paper/1905.02244/) (2019)
- [Object-Centric Learning with Slot Attention](https://mlsanity.com/paper/2006.15055/) (2020)
- [Structure-measure: A New Way to Evaluate Foreground Maps](https://mlsanity.com/paper/1708.00786/) (2017)
- [Segment Anything in High Quality](https://mlsanity.com/paper/2306.01567/) (2023)
- [Enhanced-alignment Measure for Binary Foreground Map Evaluation](https://mlsanity.com/paper/1805.10421/) (2018)
- [Anabranch Network for Camouflaged Object Segmentation](https://mlsanity.com/paper/2105.09451/) (2021)
- [Conditional Random Fields as Recurrent Neural Networks](https://mlsanity.com/paper/1502.03240/) (2015)
- [Concealed Object Detection](https://mlsanity.com/paper/2102.10274/) (2021)
- [Simultaneously Localize, Segment and Rank the Camouflaged Objects](https://mlsanity.com/paper/2103.04011/) (2021)
- [Boundary IoU: Improving Object-Centric Image Segmentation Evaluation](https://mlsanity.com/paper/2103.16562/) (2021)
- [Pixel-Adaptive Convolutional Neural Networks](https://mlsanity.com/paper/1904.05373/) (2019)
- [CascadePSP: Toward Class-Agnostic and Very High-Resolution Segmentation via Global and Local Refinement](https://mlsanity.com/paper/2005.02551/) (2020)
- [FocalClick: Towards Practical Interactive Image Segmentation](https://mlsanity.com/paper/2204.02574/) (2022)
- [Segmenting Transparent Objects in the Wild](https://mlsanity.com/paper/2003.13948/) (2020)
- [SegFix: Model-Agnostic Boundary Refinement for Segmentation](https://mlsanity.com/paper/2007.04269/) (2020)
- [Weakly-Supervised Concealed Object Segmentation with SAM-based Pseudo Labeling and Multi-scale Feature Grouping](https://mlsanity.com/paper/2305.11003/) (2023)
- [Superpixel Sampling Networks](https://mlsanity.com/paper/1807.10174/) (2018)
- [SegRefiner: Towards Model-Agnostic Segmentation Refinement with Discrete Diffusion Process](https://mlsanity.com/paper/2312.12425/) (2023)
- [RUN: Reversible Unfolding Network for Concealed Object Segmentation](https://mlsanity.com/paper/2501.18783/) (2025)
- [Fast, Exact and Multi-Scale Inference for Semantic Image Segmentation with Deep Gaussian CRFs](https://mlsanity.com/paper/1603.08358/) (2016)
- [Higher Order Conditional Random Fields in Deep Neural Networks](https://mlsanity.com/paper/1511.08119/) (2015)
- [SAMRefiner: Taming Segment Anything Model for Universal Mask Refinement](https://mlsanity.com/paper/2502.06756/) (2025)
- [Convolutional CRFs for Semantic Segmentation](https://mlsanity.com/paper/1805.04777/) (2018)
- [Regularized Frank-Wolfe for Dense CRFs: Generalizing Mean Field and Beyond](https://mlsanity.com/paper/2110.14759/) (2021)
- [Nested Unfolding Network for Real-World Concealed Object Segmentation](https://mlsanity.com/paper/2511.18164/) (2025)
- [Principled Parallel Mean-Field Inference for Discrete Random Fields](https://mlsanity.com/paper/1511.06103/) (2015)

## Abstract (arXiv)

Learned mask refiners improve segmentation accuracy, but it is hard to tell how much of the improvement comes from explicit structure rather than from extra capacity, and whether it holds up when the mask generator or its error distribution changes. UnfoldCRF treats refinement as inference in a conditional random field over pixel labels and latent region variables. Its energy has a corrected unary term, learned local pairwise interactions, and image-conditioned latent-region consistency, with a null state that lets a region with weak label agreement withdraw from the consistency term; inference unrolls damped mean-field updates on this one energy. To isolate the effect of structure, we compare against recurrent black-box refiners that read the same inputs and receive the same parameter budget, stage count, and supervision. On COD10K, UnfoldCRF beats the strongest matched control by 1.0 Fωβ point, improves all four COD metrics, and lowers the fraction of images made worse from 11.7% to 8.5%. Under a train-once protocol over five datasets and several mask sources, the 2.6M-parameter variant gains 4.2 mean DeltaIoU against 2.0 for its control, and a variant built on frozen DINOv2 features matches the strongest foundation-model refiner with about a seventh of its resident parameters while staying ahead of its own control. On mask generators never seen in training, the gain is 2.0 Fωβ points against 0.6 for the control. Zeroing individual messages shows where the corrections come from: the pairwise messages mostly fix boundaries, the region messages mostly fix non-boundary errors. Code and supporting materials will be publicly released.