cutcutcodec.core.nn.dataset.video¶
A video dataset.
Classes
|
A specific dataset to manage sub videos. |
Details
- class cutcutcodec.core.nn.dataset.video.VideoDataset(root: Path | str | bytes, shape: tuple[Integral, Integral] | None = None, *, dataaug: Callable[[Tensor], Tensor] | None = None, **kwargs)[source]
A specific dataset to manage sub videos.
Initialise and create the class.
Parameters¶
- rootpathlike
Transmitted to
Datasetinitialisator.- shapetuple[int, int], optional
If given, the video will be truncated and reshape at the shape (height, width). If not provided, the returned shape is the shape of the original video.
- dataaugcallable, optional
If provided, the function is called for each brut readed frames batch (h, w, 3*n).
- n_minint, default = 1
The minimum number of frames
- n_maxint, default = 1024
The maximum number of frames
- sizeint, default = 104857600
The targeted video size in bytes. By default 100 Mio.
- **kwargsdict
Transmitted to
Datsetinitialisator.