towhee.models.retina_face.utils.decode_landm

towhee.models.retina_face.utils.decode_landm(pre, priors, variances)[source]

Decode landm from predictions using priors to undo the encoding we did for offset regression at train time.

Parameters:

  • pre (torch.FloatTensor) – landm predictions for loc layers, Shape: [num_priors,10].

  • priors (torch.FloatTensor) – Prior boxes in center-offset form. Shape: [num_priors,4].

  • variances (list[float]) – Variances of priorboxes.

Returns:

(torch.FloatTensor)

decoded landm predictions.