Category: Research and Development

Controlling artificial intelligence poses a significant challenge in the technology world.

RealKIE, a benchmark of five challenging datasets, aims to advance key information extraction methods for enterprise applications. The diverse range of datasets offers a comprehensive testing ground for these methods.

The fundamental elements of the Orb’s software have been released on GitHub, with a dual MIT/Apache 2.0 license.

The PyTorch implementation of “ASTRA: An Action Spotting TRAnsformer for Soccer Videos”, presented at ACM MMSports 2023, has achieved third place in the SoccerNet Action Spotting Challenge 2023. This implementation, developed by arturxe2, showcases the effectiveness of ASTRA in accurately identifying and localizing actions within soccer videos.

Research on linear attention has gained significant momentum in a bid to reduce the computational load of Transformers. However, improvement strategies for attention mechanisms continue to be developed.

Temporal grounding in videos, known as VTG, is a complex challenge in video comprehension, identifying pertinent segments in untrimmed videos based on natural language prompts. Current VTG models are…

Incorporating AI-powered features, Microsoft aims to revolutionize the Xbox experience.

In 2023, AWS revealed an enhanced partnership with Hugging Face to advance customers’ AI journey. Founded in 2016, Hugging Face is a leading AI platform.

Anthropic, an AI safety and research firm, is dedicated to creating dependable, understandable, and controllable AI systems.

The Industry Documents Library (IDL) is a dataset of documents from UCSF, containing 19 million pages in various formats like PDF and TIFF.