First-Person Hand Action Benchmark

The First-Person Hand Action Benchmark is a dataset created by Garcia-Hernando et al. for hand action recognition. It consists of shoulder-mounted camera recordings of six subjects each performing 45 action classes four to six times.