Cross-view attention for visual perception tasks using multiple camera inputs

An apparatus for processing image data, the apparatus comprising:

• at least one memory;
• and
• at least one processor coupled to the at least one memory and configured to:
• obtain a plurality of input images associated with a plurality of different spatial views, wherein the plurality of input images includes a first subset of input images associated with a first group of spatial views having at least a partial overlap and a second subset of input images associated with a second group of spatial views having at least a partial overlap;
• generate a set of features based on the plurality of input images;
• generate a set of projected features based on the set of features, wherein an embedding size associated with the set of projected features is smaller than an embedding size associated with the set of features, and wherein the set of projected features includes a first subset of projected features generated based on the first subset of input images and a second subset of projected features generated based on the second subset of input images;
• and
• determine, using the set of projected features, a cross-view attention associated with the plurality of different spatial views, wherein a first cross-view attention associated with the first group of spatial views is determined using the first subset of projected features, and wherein a second cross-view attention associated with the second group of spatial views is determined using the second subset of projected features.