Assessing human skill levels in complex activities is a challenging problem with applications in sports, rehabilitation, and training. In this work, we present SkillFormer, a parameter-efficient architecture for unified multi-view proficiency estimation from egocentric and exocentric videos. Building on the TimeSformer backbone, SkillFormer introduces a CrossViewFusion module that fuses view-specific features using multi-head cross-attention, learnable gating, and adaptive self-calibration. We leverage Low-Rank Adaptation to fine-tune only a small subset of parameters, significantly reducing training costs. In fact, when evaluated on the EgoExo4D dataset, SkillFormer achieves state-of-the-art accuracy in multi-view settings while demonstrating remarkable computational efficiency, using 4.5x fewer trainable parameters and requiring 3.75x fewer training epochs than prior baselines. It excels in multiple structured tasks, confirming the value of multi-view integration for fine-grained skill assessment.
- SkillFormer: Unified Multi-View Video Understanding for Proficiency Estimation
- Edoardo Bianchi - Free University of Bozen-BolzanoAntonio Liotta - Free University of Bozen-Bolzano
- Eighteenth International Conference on Machine Vision (ICMV 2025), Vol.14114
- Osten W, Mamut E
- 9798902321873
- 9798902321880
- 0277-786X
- International Conference on Machine Vision (Paris, 19/10/2025–22/10/2025)
- SPIE - The International Society for Optical Engineering
14114 - SPIE
Washington - 9798902321873
(UNIBZ)94280969
991007295439501241 - WOS:001869126900088
- 2-s2.0-105032956766
- Faculty of Engineering
- English
- Conference proceeding
- Bianchi E, Liotta A
- Editors/Supervisors: Osten W, Mamut E