Search papers, labs, and topics across Lattice.
This study introduces ReFace, a novel spatial reorganization pipeline that enhances automatic pain assessment from facial videos by dividing facial input into four spatial quadrants prior to tokenization. By evaluating this method on the AI4Pain dataset, the authors demonstrate that it achieves a remarkable 56.00% accuracy on the test set, outperforming all previous methods under the fixed benchmark protocol. The findings indicate that spatial reorganization not only maintains the same pixel budget as full-face input but also significantly boosts accuracy, with a single quadrant configuration proving competitive at a reduced computational cost.
ReFace achieves the highest reported accuracy for pain assessment from facial videos by leveraging spatial reorganization, challenging the conventional approach of treating the face as a single entity.
Automatic pain assessment from facial video remains challenging due to the spatial heterogeneity of pain-related facial cues. This study proposes ReFace, a spatial reorganization pipeline that divides facial input into four spatial quadrants before tokenization, rather than processing the entire face as a single region. Evaluated on the AI4Pain dataset, the proposed approach achieves $56.00\%$ accuracy on the test set using video only, achieving the highest reported accuracy under the fixed AI4Pain benchmark protocol among the compared methods. Notably, the four-quadrant configuration processes the same total pixel budget as the full-face input, yet achieves higher accuracy, suggesting that spatial reorganization can improve performance under the proposed tokenization design. A single quadrant region, processing just one quarter of those pixels, remains competitive at a fraction of the computational cost.
