In the news | Tech Ninja

Included in the curated list of OpenSource AutoML frameworks, Microsoft FLAML is a robust AutoML framework with a lightweight Python library that automatically finds accurate machine learning models.

Awards | ISO.org

Gary Sullivan, Chair of the ISO/IEC JTC 1/SC 29, and Microsoft Research Principal Standards Program Manager, is part of the expert group behind the structure of multimedia files such as MP4 and 3GP.  They were honored with the prestigious Technology…

In the news | GZERO

GZERO, November 2021 | Microsoft’s Lucas Joppa says the private sector is moving beyond lofty pledges to talk about performance.

Awards | National Academy of Television Arts & Sciences

The File Format Subgroup of MPEG Systems, Working Group 3 of the ISO/IEC JTC 1/SC 29 organization led by Gary Sullivan, Principal Standards Program Manager, was recognized on by the National Academy for Television Arts and Sciences (NATAS) on November…

blindfolded businessman at edge of a cliff
Articles

We’re told to celebrate failure, because that’s how we learn – and while that’s correct, many of these failures could have been avoided. How about if we could fail before our idea reached production? Remember, that each idea that reaches…

An illustration of how the image text contrastive and translation text contrastive tasks work together to help align the space of images, English text and non-English text. On the left side of the illustration, the three domains—Image Domain, English Domain, and Non-English Domain--are segregated. An arrow labeled “Image-Captions training data” points to another depiction of the three domains where the image domain and the English domain intersect but the non-English domain is still separate and shown in gray to show that it’s not significantly affected. A two headed arrow with the label “Image-Text contrastive loss” is drawn between the image and English domains. Towards the bottom of the image, an arrow labeled “Parallel corpus training data” points to another depiction of the three domains where the English domain and the non-English domain intersect but the image domain is separate and shown in gray to indicate that it is not significantly affected. A two-headed arrow with the label “Translated Text Contrastive loss” is drawn between the English and non-English domains. Finally, a third arrow with the label “Resulting Effect” is drawn to the right of the image which points to a depiction of all three domains intersecting.
Microsoft Research Blog

| Saurabh Tiwary

Today, the Microsoft Turing team (opens in new tab) is thrilled to introduce Turing Bletchley, a 2.5-billion parameter Universal Image Language Representation model (T-UILR) that can perform image-language tasks in 94 languages. T-Bletchley has an image encoder and a universal language encoder that vectorize…

In the news | MSR Tech Minutes: Future of Haptics in Virtual Reality

Rendering haptics, or touch-sensing, in virtual reality and mixed reality (VR/MR) is crucial for rich, immersive experiences. Discover what Microsoft researchers believe to be the future of haptics in VR. Presented by Eyal Ofek (Principal Researcher) and Mike Sinclair (Senior…

2021 MSRA fellowship
Articles

The 2021 Microsoft Research Asia Fellowship has attracted 157 outstanding PhD students from 50 top universities and institutions across Asia to apply. The applicants' research fields cover in computational science, hardware and software systems, human and machine intelligence, and perception,…

ACAV100M text on top of a series of small images.
Microsoft Research Blog

| Yale Song

The natural association between visual observations and their corresponding sounds has exhibited powerful self-supervision signals for learning video representations, which makes the ever-growing amount of online video an attractive data source for self-supervised learning. However, online videos often provide imperfectly…