Balancing Biases and Preserving Privacy on Balanced Faces in the Wild.

There are demographic biases present in current facial recognition (FR) models. To measure these biases across different ethnic and gender subgroups, we introduce our Balanced Faces in the Wild (BFW) dataset. This dataset allows for the characterization of FR performance per subgroup. We found that relying on a single score threshold to differentiate between genuine and imposters sample pairs leads to suboptimal results. Additionally, performance within subgroups often varies significantly from the global average. Therefore, specific error rates only hold for populations that match the validation data. To mitigate imbalanced performances, we propose a novel domain adaptation learning scheme that uses facial features extracted from state-of-the-art neural networks. This scheme boosts the average performance and preserves identity information while removing demographic knowledge. Removing demographic knowledge prevents potential biases from affecting decision-making and protects privacy by eliminating demographic information. We explore the proposed method and demonstrate that subgroup classifiers can no longer learn from features projected using our domain adaptation scheme. For access to the source code and data, please visit https://github.com/visionjo/facerec-bias-bfw.

Cortical Tension Allocates the First Inner Cells of the Mammalian Embryo.

Every cell in our body originates from the pluripotent inner mass of the embryo, yet it is unknown how biomechanical forces allocate inner cells in vivo. Here we discover subcellular heterogeneities in tensile forces, generated by actomyosin cortical networks, which drive apical constriction to position the first inner cells of living mouse embryos. Myosin II accumulates specifically around constricting cells, and its disruption dysregulates constriction and cell fate. Laser ablations of actomyosin networks reveal that constricting cells have higher cortical tension, generate tension anisotropies and morphological changes in adjacent regions of neighboring cells, and require their neighbors to coordinate their own changes in shape. Thus, tensile forces determine the first spatial segregation of cells during mammalian development. We propose that, unlike more cohesive tissues, the early embryo dissipates tensile forces required by constricting cells via their neighbors, thereby allowing confined cell repositioning without jeopardizing global architecture.