Flat U-Net: An Efficient Ultralightweight Model for Solar Filament Segmentation in Full-disk H$α$ Images

TL;DR

Flat U-Net tackles the need for real-time, lightweight solar filament segmentation in full-disk Hα images by introducing ultralightweight SCA- and CSA-based blocks arranged in a flattened U-Net. The method preserves channel consistency across the network to dramatically reduce parameters (around 1 MB or less) while reconstructing interchannel feature information to maintain segmentation quality. Key findings show that pure SCA blocks yield high precision (≈0.93) with DSC ≈0.76 and recall ≈0.64, and introducing CSA blocks boosts DSC and recall with only modest precision trade-offs; the model remains highly compact (e.g., 0.25 MB at C=32) and robust to cloud and limb-darkening effects. The work provides open-source data, models, and code, highlighting practical deployment advantages for both ground-based and space-borne solar observation systems and suggesting broad applicability to similar astronomical segmentation tasks.

Abstract

Solar filaments are one of the most prominent features observed on the Sun, and their evolutions are closely related to various solar activities, such as flares and coronal mass ejections. Real-time automated identification of solar filaments is the most effective approach to managing large volumes of data. Existing models of filament identification are characterized by large parameter sizes and high computational costs, which limit their future applications in highly integrated and intelligent ground-based and space-borne observation devices. Consequently, the design of more lightweight models will facilitate the advancement of intelligent observation equipment. In this study, we introduce Flat U-Net, a novel and highly efficient ultralightweight model that incorporates simplified channel attention (SCA) and channel self-attention (CSA) convolutional blocks for the segmentation of solar filaments in full-disk H$α$ images. Feature information from each network layer is fully extracted to reconstruct interchannel feature representations. Each block effectively optimizes the channel features from the previous layer, significantly reducing parameters. The network architecture presents an elegant flattening, improving its efficiency, and simplifying the overall design. Experimental validation demonstrates that a model composed of pure SCAs achieves a precision of approximately 0.93, with dice similarity coefficient (DSC) and recall rates of 0.76 and 0.64, respectively, significantly outperforming the classical U-Net. Introducing a certain number of CSA blocks improves the DSC and recall rates to 0.82 and 0.74, respectively, which demonstrates a pronounced advantage, particularly concerning model weight size and detection effectiveness. The data set, models, and code are available as open-source resources.

Figures (5)

• Figure 1: The proposed Flat U-Net architecture is compared with the classical U-Net. The overall network structure of Flat U-Net, which is composed of the proposed (SCA)CSA-ConvBlock modules, optimizes interchannel weight efficiency and minimizes redundant parameters. The design presents a flattened architecture.
• Figure 2: CSA-ConvBlock.
• Figure 3: The training process of the proposed model with different umbers of channels.
• Figure 4: An example of visual comparison across different methods. (a) Full-disk H$\alpha$ solar image (without limb-darkening correction) from BBSO, captured on 2024 March 4, at 17:44:55 UT. (b) Ground truth map. (c) Segmentation result from U-Net with channel configuration [64, ..., 1024]. (d) Segmentation result from lat U-Net with channel configuration $C=32$.
• Figure 5: Examples of segmentation results without ground truth using the available model trained by $C=32$. (a) Clear full-disk H$\alpha$ solar image. (b) Full-disk H$\alpha$ solar image contaminated by clouds. (c) Segmentation results from panel (a). (d) Segmentation result from panel (b).