human archive is partnering with gig economy companies in india to collect egocentric video data for training robots. workers wear special caps with cameras to capture first-person views of everyday tasks. the startup has over 1,000 headsets deployed across various sectors, including home services and hospitality.
for indie devs, this approach offers a potential source of high-quality training data for AI models. using real-world data can improve the performance of AI in games or applications that require understanding human interactions. developers looking for diverse datasets might consider exploring partnerships or similar data collection methods.
human archive is also developing custom hardware to enhance data collection, including tactile gloves and motion capture suits. while they faced initial rejections from some companies, their unique sensor combinations may provide valuable insights for AI labs and developers.
consider experimenting with egocentric video data in your projects. capturing real-world interactions can enrich NPC behavior or improve user experience in interactive environments.