cutcutcodec.core.analysis.video.quality.uvq_google.inference.UVQInference¶

class cutcutcodec.core.analysis.video.quality.uvq_google.inference.UVQInference[source]¶

Main UVQ class.

Load all the sub modules.

forward(video: Tensor) → Tensor[source]¶

Eval the metric for each batch.

Parameters¶

videotorch.Tensor: The video slices batch of shape (n, 5, h, w, 3). It has to be in RGB in range [0, 1].

Returns¶

metrictorch.Tensor: The uvq metric for each batch of 5 frames, of shape (n,).