AGML-DenseCBAM proposes an Attention-Guided Multi-task Learning framework for robust classification of Temporomandibular Disorders using panoramic radiographs. The project explores how attention mechanisms and multi-task learning can support medical image analysis where model focus and interpretability matter.