Towards a General Model of Knowledge for Facial Analysis by Multi-Source Transfer Learning
This work addresses the need for efficient and versatile models in facial analysis, though it is incremental as it builds on existing transfer learning and distillation techniques.
The paper tackles the problem of creating general knowledge models for facial analysis by proposing a multi-source transfer learning approach that fuses existing models into a common embedding and distills them into a lightweight student model. The resulting model achieves state-of-the-art performance on 15 facial analysis tasks with 75 times fewer parameters than the teacher model.
This paper proposes a step toward obtaining general models of knowledge for facial analysis, by addressing the question of multi-source transfer learning. More precisely, the proposed approach consists in two successive training steps: the first one consists in applying a combination operator to define a common embedding for the multiple sources materialized by different existing trained models. The proposed operator relies on an auto-encoder, trained on a large dataset, efficient both in terms of compression ratio and transfer learning performance. In a second step we exploit a distillation approach to obtain a lightweight student model mimicking the collection of the fused existing models. This model outperforms its teacher on novel tasks, achieving results on par with state-of-the-art methods on 15 facial analysis tasks (and domains), at an affordable training cost. Moreover, this student has 75 times less parameters than the original teacher and can be applied to a variety of novel face-related tasks.