# Comparison of Semantic Segmentation Approaches for Horizon/Sky Line Detection

arXiv: https://arxiv.org/abs/1805.08105 · HTML: https://mlsanity.com/paper/1805.08105/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 May 2018.

## Built on by (most cited first, 4 of 4)

- [Resource Efficient Mountainous Skyline Extraction using Shallow Learning](https://mlsanity.com/paper/2107.10997/) (2021)
- [Skyline variations allow estimating distance to trees on landscape photos using semantic segmentation](https://mlsanity.com/paper/2201.08816/) (2022)
- [YUNet: Improved YOLOv11 Network for Skyline Detection](https://mlsanity.com/paper/2502.12449/) (2025)
- [HorizonNet for visual terrain navigation](https://mlsanity.com/paper/2608.30471/) (2026)

## Abstract (arXiv)

Horizon or skyline detection plays a vital role towards mountainous visual geo-localization, however most of the recently proposed visual geo-localization approaches rely on user-in-the-loop skyline detection methods. Detecting such a segmenting boundary fully autonomously would definitely be a step forward for these localization approaches. This paper provides a quantitative comparison of four such methods for autonomous horizon/sky line detection on an extensive data set. Specifically, we provide the comparison between four recently proposed segmentation methods; one explicitly targeting the problem of horizon detectionciteAhmad15, second focused on visual geo-localization but relying on accurate detection of skyline citeSaurer16 and other two proposed for general semantic segmentation -- Fully Convolutional Networks (FCN) citeLong15 and SegNetciteBadrinarayanan15. Each of the first two methods is trained on a common training set citeBaatz12 comprised of about 200 images while models for the third and fourth method are fine tuned for sky segmentation problem through transfer learning using the same data set. Each of the method is tested on an extensive test set (about 3K images) covering various challenging geographical, weather, illumination and seasonal conditions. We report average accuracy and average absolute pixel error for each of the presented formulation.