The objective of this paper is a neural network model that controls the pose and expression of a given face, using another face or modality (e.g. audio). This model can then be used for lightweight, sophisticated video and image editing. We make the following three contributions. First, we introduce a network, X2Face, that can control a source face (specified by one or more frames) using another face in a driving frame to produce a generated frame with the identity of the source frame but the pose and expression of the face in the driving frame. Second, we propose a method for training the network fully self-supervised using a large collection of video data. Third, we show that the generation process can be driven by other modalities, such as audio or pose codes, without any further training of the network. The generation results for driving a face with another face are compared to state-of-the-art self-supervised/supervised methods. We show that our approach is more robust than other methods, as it makes fewer assumptions about the input data. We also show examples of using our framework for video face editing.* Denotes equal contribution.The source face is instantiated from a single or multiple source frames, which are extracted from the same face track. The driving vector may come from multiple modalities: a driving frame from the same or another video face track, pose information, or audio information; this is illustrated in Fig. 1. The generated frame resulting from X2Face has the identity, hairstyle, etc. of the source face but the properties of the driving vector (e.g. the given pose, if pose information is given; or the driving frame's expression/pose, if a driving frame is given). The network is trained in a self-supervised manner using pairs of source and driving frames. These frames are input to two subnetworks: the embedding network and the driving network (see Fig. 2). By controlling the information flow in the network architecture, the model learns to factorise the problem. The embedding network learns an embedded face representation for the source face -effectively face frontalisation; the driving network learns how to map from this embedded face representation to the generated frame via an embedding, named the driving vector. The X2Face network architecture is described in Section 3.1, and the self-supervised training framework in Section 3.2. In addition we make two further contributions. First, we propose a method for linearly regressing from a set of labels (e.g. for head pose) or features (e.g. from audio) to the driving vector; this is described in Section 4. The performance is evaluated in Section 5, where we show (i) the robustness of the generated results compared to state-of-the-art self-supervised [45] and supervised [1] methods; and (ii) the controllability of the network using other modalities, such as audio or pose. The second contribution, described in Section 6, shows how the embedded face representation can be used for video face editing, e.g. adding facial decorations in the manne...

Interlaboratory studies in measurement science, including key comparisons, and meta-analyses in several fields, including medicine, serve to intercompare measurement results obtained independently, and typically produce a consensus value for the common measurand that blends the values measured by the participants.Since interlaboratory studies and meta-analyses reveal and quantify differences between measured values, regardless of the underlying causes for such differences, they also provide so-called 'top-down' evaluations of measurement uncertainty.Measured values are often substantially over-dispersed by comparison with their individual, stated uncertainties, thus suggesting the existence of yet unrecognized sources of uncertainty (dark uncertainty). We contrast two different approaches to take dark uncertainty into account both in the computation of consensus values and in the evaluation of the associated uncertainty, which have traditionally been preferred by different scientific communities. One inflates the stated uncertainties by a multiplicative factor. The other adds laboratory-specific 'effects' to the value of the measurand.After distinguishing what we call recipe-based and model-based approaches to data reductions in interlaboratory studies, we state six guiding principles that should inform such reductions. These principles favor model-based approaches that expose and facilitate the critical assessment of validating assumptions, and give preeminence to substantive criteria to determine which measurement results to include, and which to exclude, as opposed to purely statistical considerations, and also how to weigh them.Following an overview of maximum likelihood methods, three general purpose procedures for data reduction are described in detail, including explanations of how the consensus value and degrees of equivalence are computed, and the associated uncertainty evaluated: the DerSimonian-Laird procedure; a hierarchical Bayesian procedure; and the Linear Pool. These three procedures have been implemented and made widely accessible in a Web-based application (NIST Consensus Builder).We illustrate principles, statistical models, and data reduction procedures in four examples: (i) the measurement of the Newtonian constant of gravitation; (ii) the measurement of the halflives of radioactive isotopes of caesium and strontium; (iii) the comparison of two alternative treatments for carotid artery stenosis; and (iv) a key comparison where the measurand was the calibration factor of a radio-frequency power sensor.

No abstract