Search papers, labs, and topics across Lattice.
This paper introduces AffectFlow-DINO, a multi-task learning system designed for affect estimation that incorporates a conditional rectified-flow head to address the inherent ambiguity in facial behavior. By generating a conditional distribution rather than a single prediction, the model enables uncertainty-aware predictions across multiple tasks, including valence-arousal estimation, facial expression classification, and Action Unit detection. The results demonstrate significant improvements in performance, particularly in valence-arousal estimation and rare class detection, achieving a multi-task performance metric of 1.177, far exceeding the baseline of 0.45.
Uncertainty-aware predictions in affect estimation can dramatically enhance performance on challenging tasks, with AffectFlow-DINO achieving a multi-task performance score more than double the baseline.
We present \textbf{AffectFlow-DINO}, a multi-task learning system for the 11th ABAW challenge that extends a standard deterministic architecture with a conditional rectified-flow head to model the inherent ambiguity of in-the-wild facial behavior. Instead of predicting a single affect estimate, the model learns a conditional generative distribution, enabling uncertainty-aware one-to-many predictions through Monte Carlo sampling. The system jointly estimates continuous valence-arousal, classifies eight facial expressions, and detects twelve Action Units from static face images. Built on a frozen DINOv3 ViT-S/16 backbone, extensive ablation studies show that rectified-flow decoding consistently improves deterministic prediction, particularly for valence-arousal estimation (CCC-V $+0.058$). We further show that post-hoc threshold calibration effectively recovers performance on severely imbalanced rare classes (e.g., Fear: $3.8\% \rightarrow 33.1\%$) without retraining. Combined with backbone fine-tuning and flow retuning, the final model achieves $\mathbf{P_{MTL}=1.177}$, substantially outperforming the official challenge baseline of $P_{MTL}=0.45$.
