Accurate segmentation of wounds and scale markers in clinical images remains a significant challenge, crucial for effective wound management and automated assessment. In this study, we propose a novel dual-attention U-Net++ architecture, integrating channel-wise (SCSE) and spatial attention mechanisms to address severe class imbalance and variability in medical images effectively. Initially, extensive benchmarking across diverse architectures and encoders via 5-fold cross-validation identified EfficientNet-B7 as the optimal encoder backbone. Subsequently, we independently trained two class-specific models with tailored preprocessing, extensive data augmentation, and Bayesian hyperparameter tuning (WandB sweeps). The final model ensemble utilized Test Time Augmentation to further enhance prediction reliability. Our approach was evaluated on a benchmark dataset from the NBC 2025 & PCBBE 2025 competition. Segmentation performance was quantified using a weighted F1-score (75% wounds, 25% scale markers), calculated externally by competition organizers on undisclosed hardware. The proposed approach achieved an F1-score of 0.8640, underscoring its effectiveness for complex medical segmentation tasks.