pytorch3d.implicitron.models.feature_extractor.feature_extractor

feature_extractor

class pytorch3d.implicitron.models.feature_extractor.feature_extractor.FeatureExtractorBase(*args, **kwargs)[source]

Bases: ReplaceableBase, Module

Base class for an extractor of a set of features from images.

get_feat_dims() → int[source]
Returns:

total number of feature dimensions of the output. (i.e. sum_i(dim_i))

forward(imgs: Tensor | None, masks: Tensor | None = None, **kwargs) → Dict[Any, Tensor][source]
Parameters:

  • imgs – A batch of input images of shape (B, 3, H, W).

  • masks – A batch of input masks of shape (B, 3, H, W).

Returns:

out_feats –

A dict {f_i: t_i} keyed by predicted feature names f_i

and their corresponding tensors t_i of shape (B, dim_i, H_i, W_i).