-
公开(公告)号:US20220301097A1
公开(公告)日:2022-09-22
申请号:US17832366
申请日:2022-06-03
Applicant: Intel Corporation
Inventor: Anthony Rhodes , Ke Ding , Neil Yeung
Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed to implement dual attention vision transformers for interactive image segmentation. In some examples, the apparatus includes memory, instructions, and processor circuitry to execute and/or instantiate the instructions to partition information in an input tensor into a plurality of tensor crops, the input tensor representing image data of an image. The processor circuitry is further to execute and/or instantiate the instructions to create a dual-attention tensor representation of the input tensor, including an inter-attention tensor representation describing a first correlation between the plurality of tensor crops and an intra-attention tensor representation describing a second correlation between a plurality of layers in a tensor crop of the plurality of tensor crops.