The world is being quietly rearranged by people who write very long documents.


The title they went with Hierarchical and Multimodal Data for Daily Activity Understanding Noisy translates that to

Human activity data now includes 20 sensors and 3 levels of detail


Researchers have built a new dataset that records human activity using 20 different sensors. This data is annotated at three levels, from broad tasks down to specific steps, allowing for more detailed analysis of how people move and act.
This new dataset, DARai, offers a much richer picture of human behavior than previously available. By combining data from cameras, radar, wearables, and even insole sensors, it allows AI models to understand actions at a much finer grain. This could lead to AI systems that better understand and interact with people in real-world settings, from robotics to healthcare.
Watch whether AI models trained on this dataset show a measurable improvement in predicting human actions in complex, unscripted environments compared to models trained on less detailed data.

If you insist
Read the original →