dgs.models.engine.dgs_engine.DGSEngine.get_data

DGSEngine.get_data(ds: State) → list[torch.Tensor]

Use the similarity models of the DGS module to obtain the similarity data of the current detections.

For the similarity engine, the data consists of a list of all the input data for the similarities. This means, that for the visual similarity, the embedding is returned, and for the IoU or OKS similarities, the bbox and key point data is returned. The get_data() function will be called twice, once for the current time-step and once for the previous.