RESEARCH

OPEN SOURCE

Dynatask, a new paradigm of AI benchmarking is now available for the AI community

Today, we’re unlocking Dynabench’s full capabilities for the AI community — AI researchers can now create their own custom tasks to better evaluate the performance of NLP models in more.

September 21, 2021

RESEARCH

OPEN SOURCE

Dynatask, a new paradigm of AI benchmarking is now available for the AI community

Today, we’re unlocking Dynabench’s full capabilities for the AI community — AI researchers can now create their own custom tasks to better evaluate the performance of NLP models in more.

September 21, 2021

RESEARCH

New Captum version features more ways to build AI responsibly

September 21, 2021

RESEARCH

NLP

Textless NLP: Generating expressive speech from raw audio

September 09, 2021

Blog Posts

Blog

September 21, 2021

RESEARCH

OPEN SOURCE

Dynatask, a new paradigm of AI benchmarking is now available for the AI community

Today, we’re unlocking Dynabench’s full capabilities for the AI community — AI researchers can now create their own custom tasks to better evaluate the performance of NLP models in more.

September 21, 2021

Learn More

Blog

September 21, 2021

RESEARCH

New Captum version features more ways to build AI responsibly

We are excited to release a new version of Captum ("comprehension" in Latin), a powerful, easy-to-use model interpretability library for PyTorch. Captum 0.4 adds a variety of new functionality for understanding AI models.

September 21, 2021

Learn More

Blog

September 20, 2021

RESEARCH

COMPUTER VISION

Building AI that can generate images of things it has never seen before

We’re sharing details on how we built a new image generation model called Instance-Conditioned GAN. Unlike traditional GANs, IC-GANs can create high-quality images of entirely unfamiliar objects and scenes that aren’t part of the model’s training data.

September 20, 2021

Learn More

Blog

September 09, 2021

RESEARCH

NLP

Textless NLP: Generating expressive speech from raw audio

We’re introducing GSLM, the first language model that breaks free completely of the dependence on text for training. This “textless NLP” approach learns to generate expressive speech using only raw audio recordings as input.

September 09, 2021

Learn More

Blog

September 02, 2021

COMPUTER VISION

OPEN SOURCE

Open-sourcing Common Objects in 3D, a large-scale data set for 3D reconstruction

Facebook AI is releasing Common Objects in 3D (CO3D), a new data set of nearly 19,000 videos capturing objects along with highly accurate 3D reconstructions.

September 02, 2021

Learn More

Blog

August 26, 2021

RESEARCH

NLP

Using using pre-trained language models for database queries on any unstructured data set

We've developed a new, scalable way to search unstructured databases using natural language in lieu of traditional structured query language, such as SQL. This makes it possible to quickly search large-scale databases for answers that require reasoning over multiple facts.

August 26, 2021

Learn More

Blog

August 13, 2021

RESEARCH

COMPUTER VISION

Introducing Unidentified Video Objects, a new benchmark for open-world object segmentation

We are sharing Unidentified Video Objects (UVO), a new benchmark to facilitate research in open-world segmentation, an important computer vision task that aims to detect, segment, and track all objects exhaustively in a video.

August 13, 2021

Learn More

Blog

August 11, 2021

RESEARCH

COMPUTER VISION

Multiscale Vision Transformers: An architecture for modeling visual data

Facebook AI is sharing Multiscale Vision Transformers (MViT), a family of visual recognition models that, for the first time, incorporate the seminal concept of hierarchical representations into the powerful Transformer architecture.

August 11, 2021

Learn More

Blog

August 05, 2021

RESEARCH

Facebook AI wins first place at annual multilingual speech translation competition

We're sharing details about how we built our state of the art speech translation system

August 05, 2021

Learn More

Blog

August 02, 2021

RESEARCH

NLP

VoxPopuli: The largest open multilingual speech corpus for AI translation and more

To help NLP researchers create advanced translation and speech recognition systems for more languages spoken around the world, Facebook AI is releasing VoxPopuli. This large-scale data set provides 400K hours of unlabeled speech data in 23 languages.

August 02, 2021

Learn More

Blog

July 30, 2021

RESEARCH

Introducing droidlet, a one-stop shop for modularly building intelligent agents

We are open-sourcing droidlet, a “one-stop shop” platform for AI robotics research. Droidlet makes it easy to mix and match different computer vision and language models to build versatile, intelligent agents that can operate in simulation or in the real world.

July 30, 2021

Learn More

Blog

July 18, 2021

RESEARCH

COMPUTER VISION

Introducing few-shot neural architecture search

We’re sharing our work on few-shot neural architecture search (NAS), which combines the accuracy of vanilla NAS with the speed and efficiency of one-shot NAS. Few-shot NAS lets anyone design a powerful custom model quickly, with just a few GPUs.