# SparseMask: Differentiable Connectivity Learning for Dense Image Prediction

arXiv: https://arxiv.org/abs/1904.07642 · HTML: https://mlsanity.com/paper/1904.07642/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Apr 2019.

## Built on by (most cited first, 7 of 7)

- [OpenEarthMap: A Benchmark Dataset for Global High-Resolution Land Cover Mapping](https://mlsanity.com/paper/2210.10732/) (2022)
- [Fantastic Weights and How to Find Them: Where to Prune in Dynamic Sparse Training](https://mlsanity.com/paper/2306.12230/) (2023)
- [Neural Architecture Search for Dense Prediction Tasks in Computer Vision](https://mlsanity.com/paper/2202.07242/) (2022)
- [Real-Time Semantic Segmentation: A Brief Survey & Comparative Study in Remote Sensing](https://mlsanity.com/paper/2309.06047/) (2023)
- [Neural Architecture Search generated Phase Retrieval Net for Real-time Off-axis Quantitative Phase Imaging](https://mlsanity.com/paper/2210.14231/) (2022)
- [The devil is in discretization discrepancy. Robustifying Differentiable NAS with Single-Stage Searching Protocol](https://mlsanity.com/paper/2405.16610/) (2024)
- [More Than Where You Are: Learning Semantics, Structure, and Geometry from Cross-View Localization](https://mlsanity.com/paper/2607.12429/) (2026)

## Abstract (arXiv)

In this paper, we aim at automatically searching an efficient network architecture for dense image prediction. Particularly, we follow the encoder-decoder style and focus on designing a connectivity structure for the decoder. To achieve that, we design a densely connected network with learnable connections, named Fully Dense Network, which contains a large set of possible final connectivity structures. We then employ gradient descent to search the optimal connectivity from the dense connections. The search process is guided by a novel loss function, which pushes the weight of each connection to be binary and the connections to be sparse. The discovered connectivity achieves competitive results on two segmentation datasets, while runs more than three times faster and requires less than half parameters compared to the state-of-the-art methods. An extensive experiment shows that the discovered connectivity is compatible with various backbones and generalizes well to other dense image prediction tasks.