September 2020 Summaries

In the blog post, Kanstantsin Sokal from the MediaPipe team introduces a new face transform estimation module that enhances the existing Face Mesh solution, capable of estimating a 3D face shape in real-time on mobile devices through 468 landmarks. The module offers a semantic API for determining 3D face poses and attaching virtual objects, such as glasses or masks, by transitioning from screen coordinate space to a metric 3D space. This transition is facilitated by Procrustes Analysis, which is lightweight and operates on a CPU with minimal resource use. The module enables the alignment of virtual 3D objects with detected faces by using a canonical face model and a face pose transformation matrix. It includes a face effect renderer that supports real-time rendering of 3D objects and face mesh effects via OpenGL ES 2.0, which can be utilized on mobile applications. The module is part of the MediaPipe Face Mesh solution and includes examples and resources for developers to implement custom use cases.

DevFest 2020, held virtually from October 16-18, brought together thousands of developers globally for a weekend of community-led learning on Google technologies. Traditionally an in-person series, this year's virtual format allowed developers to connect during a time when community interaction was highly sought after. The event featured diverse speakers, including young developers, self-taught programmers, and leaders from startups, alongside Google representatives and Women Techmakers. Hosted by Google Developer Groups, the sessions covered a wide range of technologies such as Android, Google Cloud Platform, TensorFlow, and Flutter. The event emphasized the importance of grassroots community efforts in advancing technology, with Google providing support through speakers and custom technical content. Despite the challenges of 2020, the enthusiasm for DevFest remained strong as developers sought new opportunities for collaboration and learning.

The Google Nest Device Access Console, launched on September 22, 2020, allows both individuals and qualified partners to securely access and control Nest products via their apps and solutions through the Smart Device Management API. Early Access Program participants such as Alarm.com, Control4, and Vivint Smart Home have successfully integrated Nest devices like thermostats, cameras, and doorbells. The program emphasizes privacy and security, requiring partners to submit qualified use cases and complete security assessments before commercial use, ensuring data protection and user consent for device management. It also offers a sandbox development environment for both commercial developers and individual enthusiasts to test and integrate Nest devices, promoting a flexible and secure smart home ecosystem.

Google Pay's engineering team chose Flutter to advance their global product development, initially driven by the success of Google Pay India, formerly known as Tez. The original vision was to provide a simple and secure payment app, which expanded to include features like bill payments, phone recharges, instant loans, and ticket purchases. As the team aimed to scale globally, they needed a development environment that allowed efficient deployment across both iOS and Android platforms, leading them to select Flutter for its ability to write code once in Dart and deploy it on multiple operating systems. This choice was bolstered by Flutter's Just-in-Time compiler, hot reload, and high-performance deployment capabilities. Starting with a small team, they focused on a complete rewrite rather than a hybrid approach to ensure scalability and maintainability. After months of development, Google Pay's Flutter implementation entered open beta in India and Singapore, with plans to extend globally. The transition is significant, marking Google Pay as one of the largest deployments on the Flutter platform, thanks to the continued advancements by the Flutter team.

Coral has introduced the M.2 Accelerator with Dual Edge TPU, enhancing machine learning performance and expanding the capabilities of edge computing by enabling multi-Edge TPU configurations for processing sophisticated models. This accelerator is being utilized in Google's Series One room kits for Google Meet, where it improves audio clarity by employing TrueVoice® technology for noise cancellation, ensuring clear communication even in challenging acoustic environments. To make edge AI more accessible, Coral has reduced prices on several products, including prototyping devices and sensors, and expanded their System-on-Module offerings. Additionally, they highlight innovative uses of their technology, such as a pose-based AI Director tool for multi-camera video production, encouraging developers to share projects and feedback.

Google for Startups Accelerator has announced the opening of applications for its third accelerator class in Japan, starting January 2021, with a focus on established startups addressing significant social and environmental issues. Applications are open until October 30, 2020, for companies that contribute to the Japanese economy by tackling challenges such as an aging society, energy sustainability, and rural revitalization. The program, designed for later-stage companies with a strong market fit and growth potential, offers tailored mentorship, product advice, technical workshops, and networking opportunities with other founders, venture capitalists, and industry experts. Selected startups will join a global alumni network of over 500 companies, benefiting from Google's resources to scale their businesses effectively.

G Suite, widely recognized as a suite of productivity and communication applications, also serves as a robust developer platform allowing for custom integrations and enhancements. It provides an extensible foundation with components like G Suite Add-ons, Chatbots, and a comprehensive library of REST APIs, which facilitate the creation of custom experiences and data interactions across its apps. Developers can leverage tools like Google Cloud, BigQuery, and Connected Sheets to augment G Suite’s capabilities, enabling sophisticated data analysis and the integration of machine learning through familiar interfaces. Additionally, G Suite supports diverse development preferences with tools such as Apps Script for quick solution development and Alternate Runtimes for greater flexibility in coding environments. The platform is backed by robust security features and administrative tools, allowing developers to confidently deploy and distribute their solutions via the G Suite Marketplace.