A repository of multi-task models MTCS. The complete training and validation code will be released after the paper is accepted.
While MTCS does not surpass P2PNet in counting accuracy, it achieves results that are highly comparable.
Due to the robustness gained from multi-task learning, MTCS outperformed YOLOv11 in segmentation accuracy.
Taking Photovoltaic Disaster Damage Assessment as an Example. Areas are categorized into different damage levels based on the number of pits: green represents 0 pits, blue represents 1–2 pits, yellow represents 3–5 pits, and red represents more than 5 pits.
