plot_videos#

deepinv.utils.plot_videos(vid_list, titles=None, time_dim=2, rescale_mode='min_max', display=False, figsize=None, dpi=None, save_fn=None, return_anim=False, anim_writer=None, anim_kwargs={}, **plot_kwargs)[source]#

Plots and animates a list of image sequences.

Plots videos as sequence of side-by-side frames, and saves animation (e.g. GIF) or displays as interactive HTML in notebook. This is useful for e.g. time-varying inverse problems. Individual frames are plotted with deepinv.utils.plot()

Plots videos as sequence of side-by-side frames, and saves animation (e.g. GIF) or displays as interactive HTML in notebook. This is useful for e.g. time-varying inverse problems. Individual frames are plotted with deepinv.utils.plot() vid_list can either be a video or a list of them. A video is defined as images of shape (B,C,H,W) augmented with a time dimension specified by time_dim, e.g. of shape (B,C,T,H,W) and time_dim=2. All videos must be same time-length.

Per frame of the videos, this function calls deepinv.utils.plot(), see its params to see how the frames are plotted.

To display an interactive HTML video in an IPython notebook, use display=True. Note IPython must be installed for this. Per frame of the videos, this function calls deepinv.utils.plot(), see its params to see how the frames are plotted. To display an interactive HTML video in an IPython notebook, use display=True. Note IPython must be installed for this.

Examples:

Display list of image sequences live in a notebook:

>>> from deepinv.utils import plot_videos
>>> x = torch.rand((1, 3, 5, 8, 8)) # B,C,T,H,W image sequence
>>> y = torch.rand((1, 3, 5, 16, 16))
>>> plot_videos([x, y], display=True) # Display interactive view in notebook (requires IPython)
>>> plot_videos([x, y], save_fn="vid.gif") # Save video as GIF
Parameters:

  • vid_list (Union[torch.Tensor, List[torch.Tensor]]) – video or list of videos as defined above.

  • titles (Union[str, List[str]]) – titles of images in frame, defaults to None.

  • time_dim (int) – time dimension of the videos. All videos should have same length in this dimension, or length 1. After indexing this dimension, the resulting images should be of shape (B,C,H,W). Defaults to 2.

  • rescale_mode (str) – rescaling mode for deepinv.utils.plot(), defaults to "min_max"

  • vid_list – video or list of videos as defined above

  • titles – titles of images in frame, defaults to None

  • time_dim – time dimension of the videos. All videos should have same length in this dimension, or length 1. After indexing this dimension, the resulting images should be of shape (B,C,H,W). Defaults to 2

  • rescale_mode – rescaling mode for deepinv.utils.plot(), defaults to "min_max"

  • display (bool) – display an interactive HTML video in an IPython notebook, defaults to False

  • figsize (tuple[int], None) – size of the figure. If None, calculated from size of img list.

  • save_fn (str) – if not None, save the animation to this filename. File extension must be provided, note anim_writer might have to be specified. Defaults to None

  • save_fn – if not None, save the animation to this filename. File extension must be provided, note anim_writer might have to be specified. Defaults to None

  • anim_writer (str) – animation writer, see https://matplotlib.org/stable/users/explain/animations/animations.html#animation-writers, defaults to None

  • return_anim (bool) – return matplotlib animation object, defaults to False

  • dpi (int) – DPI of saved videos.

  • anim_kwargs (dict) – keyword args for matplotlib FuncAnimation init

  • plot_kwargs (dict) – kwargs to pass to deepinv.utils.plot()

Examples using plot_videos:#

Tour of MRI functionality in DeepInverse

Tour of MRI functionality in DeepInverse

Self-supervised MRI reconstruction with Artifact2Artifact

Self-supervised MRI reconstruction with Artifact2Artifact