["Transactions in GIS, Volume 30, Issue 5, August 2026. ", "\nABSTRACT\nSeismic facies identification is a crucial technique for geological stratigraphic analysis. To reduce the complexity and subjectivity associated with manual interpretation, deep learning‐based algorithms have been increasingly developed for seismic facies identification. However, existing methods still face several challenges, including weak robustness, low detection accuracy for small‐scale targets, and blurred boundary segmentation. To address these issues, a multi‐scale seismic attribute feature fusion network with a Gini feedback module, termed GF‐MSAFF, is proposed. First, a multi‐scale attribute feature fusion module is designed by progressively integrating multi‐scale features derived from different seismic attributes. Second, the Gini index is used to evaluate prediction results and identify regions that have a significant influence on segmentation outcomes. Subsequently, a Gini feedback module is constructed to optimize the model and incorporate a gating mechanism for feature channel selection. Finally, a dynamic edge loss function is introduced to extract edge labels from semantic labels and dynamically adjust the labels during the training process. To evaluate the effectiveness of the proposed model, comparative experiments are conducted against classical and state‐of‐the‐art algorithms using the Parihaka seismic dataset and Netherlands F3 datasets. The experimental results demonstrate that the proposed approach captures more detailed structural and boundary information, thereby enhancing segmentation performance. Compared with existing best‐performing models, GF‐MSAFF improves MIoU, MPA, and Recall by 0.606%, 0.337%, and 0.191%, respectively, on the Parihaka seismic dataset, and by 0.436%, 0.249%,and 0.208%, respectively, on the Netherlands F3 dataset.\n"]