Local Recurrent Attention
Using local recurrent attention weights we can derive synchronization information between motion and words.
- Compositional motion
- Single action
Abstract
In this paper, we investigate building a sequence to sequence architecture for motion to language translation and synchronization. The aim is to translate motion capture inputs into English natural-language descriptions, such that the descriptions are generated synchronously with the actions performed, enabling semantic segmentation as a byproduct, but without requiring synchronized training data.
We propose a new recurrent formulation of local attention that is suited for synchronous text generation, as well as an improved motion encoder architecture better suited to smaller data and for synchronous generation. We evaluate both contributions in individual experiments, using the standard BLEU4 metric, as well as a simple semantic equivalence measure, on the KIT motion language dataset. In a follow-up experiment, we assess the quality of the synchronization of generated text in our proposed approaches through multiple evaluation metrics. We find that both contributions to the attention mechanism and the encoder architecture additively improve the quality of generated text (BLEU and semantic equivalence), but also of synchronization.
Motion Frozen in Time
We can use attention weights as transparency level of motion frames and better visualize synchronization.
Graphical abstract
Comparison
The novelty of our approach stems from the ability to infer local mapping between motion primitives and their subtitles using local recurrent attention enabling for the first time a synchronized captioning.
Segmentation process
For quantitative evaluation we define motion and language segments and use metrics such as IoU.
