ABAW 10th competition tackles emotional mimicry and violence detection at CVPR 2026

The 10th Affective & Behavior Analysis in-the-Wild (ABAW) Workshop and Competition, held at CVPR 2026, advances multimodal human-centered AI for real-world environments. The competition features six distinct challenges: continuous affect (valence-arousal) estimation, discrete affect (expression and action unit) recognition, emotional mimicry intensity estimation, ambivalence/hesitancy recognition, and fine-grained violence detection. These tasks are grounded in large-scale in-the-wild datasets, ensuring benchmarks reflect real-world complexity. The workshop maintains a dual structure with both a competition track and a paper track, fostering collaboration and innovation in affective computing and behavioral understanding.

The paper track showcases contributions spanning pose, motion, and behavior estimation, affect modeling, multimodal learning, benchmarks, datasets, evaluation protocols, fairness, robustness, and deployment. By addressing key aspects of human affect—from basic expressions to complex behaviors like violence and hesitancy—the workshop drives the development of AI systems that can interpret nuanced human states. This work is critical for applications in mental health monitoring, human-robot interaction, content moderation, and safety-critical AI. The 10th ABAW continues to serve as a key platform for benchmarking and shaping next-generation human-centered AI.