VisionNet aims to democratize access to high-quality vision datasets for robotics research and development. Today, large language models (LLMs) are trained on enormous amounts of one-dimensional data collected from the internet. This data - created and authored by humans, has been used without direct compensation to the original creators.
The core challenge in deploying effective robotic systems lies in acquiring robust, diverse, and high-quality datasets that capture real life from a first-person point of view. These datasets are fundamentally different from the text-based or static visual data available online and require new capture form factors such as smart glasses, head-mounted cameras, and other wearable devices to gather rich, real-world perspectives.
VisionNet is building a platform to collect and curate these first-person datasets to advance robotics research, with a central focus on fairly compensating the contributors of this data.
VisionNet is a part of San Francisco based robotics company Madvision, Inc.
For any information please text at x.com/madvisionhq