Generic Autodidactic Models, also known as Models Genesis, was developed by researchers at Arizona State University. They are ex nihilo (without manual labeling), self-taught (learned by self-supervision), and generic (served as source models for producing application-specific target models). Their rigorous testing shows that in all five target 3D applications, including segmentation and classification, their Models Genesis considerably exceed learning from scratch.
DiscoFaceGAN is a technique for face image creation with DISentangled, precisely-controllable latent representations for identity of hypothetical persons, expression, position, and illumination, according to research from Microsoft Research Asia. In order to train the network to mimic the picture creation of an analytical 3D face deformation and rendering process, they integrated 3D priors into adversarial learning. They added contrastive learning to encourage disentanglement by comparing pairs of generated images in order to address the generation freedom brought on by the gap in the domain between real and rendered faces. The factor variations are very thoroughly disentangled and the characteristics of a generated face can be accurately regulated, according to experiments on their imitative-contrastive learning.
Open Source News:
Assured Open Source Software (Assured OSS), a tool that aids developers in defending against supply chain security assaults by routinely scanning and analyzing some of the most well-known software libraries for vulnerabilities, was introduced by Google around a year ago. With support for more than a thousand Java and Python packages, Google today announced the general availability of Assured OSS. Although Google previously withheld pricing information when the service was first introduced, the firm has since revealed that it will be offered for free.
On Tuesday, a new client-side runtime that is open source and targeted towards web developers was made available. Web developers can utilize their preferred frontend libraries using Socket Runtime because it can be installed on any OS and be used on desktops, smartphones, and other mobile devices. Additionally, peer-to-peer networking is used to deploy apps rather than a central cloud server.