# DeepStrip: High Resolution Boundary Refinement

arXiv: https://arxiv.org/abs/2003.11670 · HTML: https://mlsanity.com/paper/2003.11670/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Mar 2020.

## Built on by (most cited first, 9 of 9)

- [Reviving Iterative Training with Mask Guidance for Interactive Segmentation](https://mlsanity.com/paper/2102.06583/) (2021)
- [Progressive Semantic Segmentation](https://mlsanity.com/paper/2104.03778/) (2021)
- [SegRefiner: Towards Model-Agnostic Segmentation Refinement with Discrete Diffusion Process](https://mlsanity.com/paper/2312.12425/) (2023)
- [Ultra-high Resolution Image Segmentation via Locality-aware Context Fusion and Alternating Local Enhancement](https://mlsanity.com/paper/2109.02580/) (2021)
- [Boosting Semantic Segmentation with Semantic Boundaries](https://mlsanity.com/paper/2304.09427/) (2023)
- [rNCA: Self-Repairing Segmentation Masks](https://mlsanity.com/paper/2512.13397/) (2025)
- [Generating Superpixels for High-resolution Images with Decoupled Patch Calibration](https://mlsanity.com/paper/2108.08607/) (2021)
- [A Robust Morphological Approach for Semantic Segmentation of Very High Resolution Images](https://mlsanity.com/paper/2208.01254/) (2022)
- [PnPNet: Pull-and-Push Networks for Volumetric Segmentation with Boundary Confusion](https://mlsanity.com/paper/2312.08323/) (2023)

## Abstract (arXiv)

In this paper, we target refining the boundaries in high resolution images given low resolution masks. For memory and computation efficiency, we propose to convert the regions of interest into strip images and compute a boundary prediction in the strip domain. To detect the target boundary, we present a framework with two prediction layers. First, all potential boundaries are predicted as an initial prediction and then a selection layer is used to pick the target boundary and smooth the result. To encourage accurate prediction, a loss which measures the boundary distance in the strip domain is introduced. In addition, we enforce a matching consistency and C0 continuity regularization to the network to reduce false alarms. Extensive experiments on both public and a newly created high resolution dataset strongly validate our approach.