interpolate_1d

behavenet.plotting.cond_ae_utils.interpolate_1d(interp_type, model, ims_0, latents_0, labels_0, labels_sc_0, mins, maxes, input_idxs, n_frames, crop_type=None, mins_sc=None, maxes_sc=None, crop_kwargs=None, marker_idxs=None, ch=0)[source]

Return reconstructed images created by interpolating through latent/label space.

Parameters:

  • interp_type (str) – ‘latents’ | ‘labels’

  • model (behavenet.models object) – autoencoder model

  • ims_0 (torch.Tensor) – base images for interpolating labels, of shape (1, n_channels, y_pix, x_pix)

  • latents_0 (np.ndarray) – base latents of shape (1, n_latents); only two of these dimensions will be changed if interp_type=’latents’

  • labels_0 (np.ndarray) – base labels of shape (1, n_labels)

  • labels_sc_0 (np.ndarray) – base scaled labels in pixel space of shape (1, n_labels, y_pix, x_pix)

  • mins (array-like) – minimum values of all labels/latents

  • maxes (array-like) – maximum values of all labels/latents

  • input_idxs (array-like) – indices of labels/latents that will be interpolated

  • n_frames (int) – number of interpolation points between mins and maxes (inclusive)

  • crop_type (str or NoneType, optional) – currently only implements ‘fixed’; if not None, cropped images are returned, and returned labels are also cropped so that they can be plotted on top of the cropped images; if None, returned cropped images are empty and labels are relative to original image size

  • mins_sc (list, optional) – min values of scaled labels that correspond to min values of labels when using conditional encoders

  • maxes_sc (list, optional) – max values of scaled labels that correspond to max values of labels when using conditional encoders

  • crop_kwargs (dict, optional) – define center and extent of crop if crop_type=’fixed’; keys are ‘x_0’, ‘x_ext’, ‘y_0’, ‘y_ext’

  • marker_idxs (list, optional) – indices of labels_sc_0 that will be interpolated; note that this is analogous but different from input_idxs, since the 2d tensor labels_sc_0 has half as many label dimensions as latents_0 and labels_0

  • ch (int, optional) – specify which channel of input images to return (can only be a single value)

Returns:

  • ims_list (list of list of np.ndarray) interpolated images

  • labels_list (list of list of np.ndarray) interpolated labels

  • ims_crop_list (list of list of np.ndarray) interpolated , cropped images

Return type:

tuple