Abstract:Deep learning methods, especially deep convolutional neural networks (DCNNs), have been widely used in high-resolution remote sensing image (HRSI) semantic segmentation. In literature, most successful DCNNs are artificially designed through a large number of experiments, which often consume lots of time and depend on rich domain knowledge. Recently, neural architecture search (NAS), as a direction for automatically designing network architectures, has achieved great success in different kinds of computer vision tasks. For HRSI semantic segmentation, NAS faces two major challenges: (1) The task's high complexity degree, which is caused by the pixel-by-pixel prediction demand in semantic segmentation, leads to a rapid expansion of the search space; (2) HRSI semantic segmentation often needs to exploit long-range dependency (i.e., a large spatial context), which means the NAS technique requires a lot of display memory in the optimization process and can be tough to converge. With the aforementioned considerations in mind, we propose a new decoupling NAS (DNAS) framework to automatically design the network architecture for HRSI semantic segmentation. In DNAS, a hierarchical search space with three levels is recommended: path-level, connection-level, and cell-level. To adapt to this hierarchical search space, we devised a new decoupling search optimization strategy to decrease the memory occupation. More specifically, the search optimization strategy consists of three stages: (1) a light super-net (i.e., the specific search space) in the path-level space is trained to get the optimal path coding; (2) we endowed the optimal path with various cross-layer connections and it is trained to obtain the connection coding; (3) the super-net, which is initialized by path coding and connection coding, is populated with kinds of concrete cell operators and the optimal cell operators are finally determined. It is worth noting that the well-designed search space can cover various network candidates and the optimization process can be done efficiently. Extensive experiments on the publicly open GID and FU datasets showed that our DNAS outperformed the state-of-the-art methods, including artificial networks and NAS methods.

Searching sharing relationship for instance segmentation decoder

Hierarchical Shared Architecture Search for Real-Time Semantic Segmentation of Remote Sensing Images

Sed: Searching Enhanced Decoder with Switchable Skip Connection for Semantic Segmentation

DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation

Automatic Network Architecture Search for RGB-D Semantic Segmentation

DNAS: Decoupling Neural Architecture Search for High-Resolution Remote Sensing Image Semantic Segmentation

Efficient Architecture Search for Real-Time Instance Segmentation.

Surrogate-assisted Multi-objective Neural Architecture Search for Real-time Semantic Segmentation

Template-Based Automatic Search of Compact Semantic Segmentation Architectures

FBNetV5: Neural Architecture Search for Multiple Tasks in One Run

A Two-Pipeline Instance Segmentation Network via Boundary Enhancement for Scene Understanding

DIY Your EasyNAS for Vision: Convolution Operation Merging, Map Channel Reducing, and Search Space to Supernet Conversion Tooling.

Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation

SP-NAS: Serial-to-Parallel Backbone Search for Object Detection

Fast Neural Architecture Search of Compact Semantic Segmentation Models Via Auxiliary Cells

Exploring Shared Structures and Hierarchies for Multiple NLP Tasks

Neural Architecture Search for Joint Human Parsing and Pose Estimation

CNASV - A Convolutional Neural Architecture Search-Train Prototype for Computer Vision Task.

Evolving blocks by segmentation for neural architecture search

Towards bi-directional skip connections in encoder-decoder architectures and beyond

SCARLET-NAS: Bridging the Gap between Stability and Scalability in Weight-sharing Neural Architecture Search