Learning Deep Representation for Face Alignment with Auxiliary Attributes

PROJECT TITLE :

Learning Deep Representation for Face Alignment with Auxiliary Attributes

ABSTRACT:

In this study, we tend to show that landmark detection or face alignment task isn't one and independent problem. Instead, its robustness will be greatly improved with auxiliary data. Specifically, we tend to jointly optimize landmark detection together with the popularity of heterogeneous however subtly correlated facial attributes, like gender, expression, and appearance attributes. This can be non-trivial since completely different attribute inference tasks have completely different learning difficulties and convergence rates. To deal with this drawback, we have a tendency to formulate a completely unique tasks-constrained deep model, which not solely learns the inter-task correlation however additionally employs dynamic task coefficients to facilitate the optimization convergence when learning multiple complicated tasks. Extensive evaluations show that the proposed task-constrained learning (i) outperforms existing face alignment methods, especially in dealing with faces with severe occlusion and pose variation, and (ii) reduces model complexity drastically compared to the state-of-the-art methods based on cascaded deep model.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here