FEEL pairs piezoresistive glove force with egocentric video

FEEL, described in a preprint posted to arXiv on March 16, pairs force readings from custom piezoresistive gloves with egocentric video, releasing what its authors call the first large-scale dataset of its kind. The corpus, the paper says, "contains approximately 3 million force-synchronized frames of natural unscripted manipulation in kitchen environments, with 45% of frames involving hand-object contact."

The eight-author team, led by Eadom Dessalene, frames force as foundational: "Because force is the underlying cause that drives physical interaction, it is a critical primitive for physical action understanding." They apply the dataset to two task families: temporal and pixel-level contact segmentation, and action representation learning where force prediction serves as a self-supervised pretraining objective for video backbones.

The abstract reports "state-of-the-art temporal contact segmentation results and competitive pixel-level segmentation results without any need for manual contacted object segmentation annotations," and says force-pretrained representations improve transfer on EPIC-Kitchens, SomethingSomething-V2, EgoExo4D, and Meccano. Two researchers on our tracker circulated the link within days of the arXiv post.

The abstract publishes no per-benchmark accuracy figures and gives no timeline for releasing the gloves or the data.