# Per-finger force estimation

Jinglianwen has released a binocular EGO+EMG multimodal hand-interaction dataset aimed at the scarcity of force and muscle-signal data for dexterous manipulation. It covers 28 domains and 1,170 real tasks, adding 16-channel surface EMG and five per-hand per-finger force channels to binocular first-person video and hand pose, with pipeline figures including under 3 ms multimodal synchronization deviation. The per-finger force is inferred by the Fusion2Force fusion model rather than measured directly, so it should be read together with the contact mask and validity flags.

Jinglianwen has released a binocular EGO+EMG multimodal hand-interaction dataset aimed at the scarcity of force and muscle-signal data for dexterous manipulation. It covers 28 domains and 1,170 real tasks, adding 16-channel surface EMG and five per-hand per-finger force channels to binocular first-person video and hand pose, with pipeline figures including under 3 ms multimodal synchronization deviation. The per-finger force is inferred by the Fusion2Force fusion model rather than measured directly, so it should be read together with the contact mask and validity flags.