Abstract: Superpixels have been widely used in computer vision tasks due to their representational and computational efficiency. Meanwhile, deep learning and end-to-end framework have made great progress in various fields including computer vision. However, existing superpixel algorithms cannot be integrated into subsequent tasks in an end-to-end way. Traditional algorithms and deep learning-based algorithms are two main streams in superpixel segmentation. The former is non-differentiable and the latter needs a non-differentiable post-processing step to enforce connectivity, which constraints the integration of superpixels and downstream tasks. In this paper, we propose a deep learning-based superpixel segmentation algorithm SIN which can be integrated with downstream tasks in an end-to-end way. Owing to some downstream tasks such as visual tracking require real-time speed, the speed of generating superpixels is also important. To remove the post-processing step, our algorithm enforces spatial connectivity from the start. Superpixels are initialized by sampled pixels and other pixels are assigned to superpixels through multiple updating steps. Each step consists of a horizontal and a vertical interpolation, which is the key to enforcing spatial connectivity. Multi-layer outputs of a fully convolutional network are utilized to predict association scores for interpolations. Experimental results show that our approach runs at about 80fps and performs favorably against state-of-the-art methods. Furthermore, we design a simple but effective loss function which reduces much training time. The improvements of superpixel-based tasks demonstrate the effectiveness of our algorithm. We hope SIN will be integrated into downstream tasks in an end-to-end way and benefit the superpixel-based community. Code is available at: \href{https://github.com/yuanqqq/SIN}{https://github.com/yuanqqq/SIN}.

IRSL: Iterative Refining Superpixel Lattice

A Single-Image Super-Resolution Method Based on Progressive-Iterative Approximation

Lazy Random Walks for Superpixel Segmentation

Fast and robust superpixel generation method

Learning Series-Parallel Lookup Tables for Efficient Image Super-Resolution

SQL: Superpixels Via Quaternary Labeling

Linear Spectral Clustering Superpixel.

High Quality Superpixel Generation Through Regional Decomposition

NSLIC: SLIC Superpixels Based on Nonstationarity Measure

Incorporating Texture into SLIC Super-pixels Method for High Spatial Resolution Remote Sensing Image Segmentation.

Image superpixel segmentation based on hierarchical multi-level LI-SLIC

SIN - Superpixel Interpolation Network.

Real-time Image Smoothing via Iterative Least Squares

Improved Simple Linear Iterative Clustering Algorithm Using Hsl Color Space

Superpixel Segmentation Using Linear Spectral Clustering

Single Image Super-Resolution via Iterative Collaborative Representation.

LatticeNet: Towards Lightweight Image Super-Resolution with Lattice Block

Image Super-Resolution with Facet Improvement and Detail Enhancement Based on Local Self Examples

An unsupervised semantic segmentation method that combines the ImSE-Net model with SLICm superpixel optimization

Rethinking Superpixel Segmentation from Biologically Inspired Mechanisms

An Attention-Based Approach for Single Image Super Resolution.