DC4Former: Orientation-Stable UAV Disaster Image Segmentation via Diagonal-Complemented C4 Consistency

UAV disaster imagery is often acquired under varying flight headings and camera yaw orientations, which can change the in-plane orientation of the observed scene. We propose DC4Former, an orientation-stable segmentation framework built on LRFormer. The architecture-only variant, C4Former, targets four grid-preserving rotations (0∘, 90∘, 180∘, and 270∘) by redesigning three rotation-sensitive operations: stride-1 spatial convolution, strided downsampling, and low-resolution pooling attention. It uses C4-symmetrized kernels, feature-rotation-averaged downsampling, and aligned four-view pooling attention with constrained branch fusion. DC4Former further adds a single 45∘ rotation augmentation during training to improve empirical robustness at diagonal angles while retaining C4 as the grid-preserving architectural basis. FloodNet and RescueNet experiments use an eight-angle protocol and a 24-angle stress test. Compared with LRFormer, DC4Former increases the three-seed worst-angle foreground mIoU from 70.14% to 76.83% on FloodNet and from 54.90% to 67.66% on RescueNet, while reducing the RescueNet eight-angle standard deviation from 4.17 to 0.33. Relative to LRFormer-RandRot, which adds no inference-time overhead, the observed deployment trade-off is dataset- and protocol-dependent. On FloodNet, LRFormer-RandRot has slightly higher All-8, Worst-8, Mean-24, and Worst-24 values. On RescueNet, it has higher Mean-24 and Worst-24 and a lower Std.-24, whereas DC4Former has higher All-8 and Worst-8 and a lower Std.-8 under the eight-angle protocol at increased inference latency. Taken together, these results show that operator-aware C4 consistency combined with diagonal training exposure provides an architecture-level route to improved reliability across the eight principal orientations relative to the unmodified LRFormer backbone.