We present a human-validated dataset that contains 224 high-resolution, multi-view video clips and audio recordings of emotionally charged interactions between eight couples of actors. The dataset is fully annotated with categorical labels for four basic emotions (anger, happiness, sadness, and surprise) and continuous labels for valence, activation, power, and anticipation provided by five annotators for each actor. In total, we recorded eight pairs of actors (three female only, two male only, three mixed), with each pair performing seven scenarios, each consisting of four subscenarios. This resulted in 224 video clips with a total length of 143 minutes or 252,457 frames. The average length of the recorded video clips in our dataset was around 38 seconds.

The data is only to be used for non-commercial scientific purposes.