Improved DeepLabv3+ for Semantic Segmentation via Multi-Scale Context Fusion and Feature Attention. JCSFT. 2026;2(3):101-113. doi:10.63313/JCSFT.9051