US Patent Issued to GOOGLE on Aug. 4 for "Channel fusion for vision-language representation learning" (New Hampshire, Colorado Inventors)

ALEXANDRIA, Va., Aug. 5 -- United States Patent no. 12,700,225, issued on Aug. 4, was assigned to GOOGLE LLC (Mountain View, Calif.).

"Channel fusion for vision-language representation learning" was invented by Anthony J. Piergiovanni (Denver) and Maxwell Mbabilla Aladago (Hanover, N.H.).

According to the abstract* released by the U.S. Patent & Trademark Office: "Provided is an approach that aligns multi-modal tokens using cross-attention without losing the advantages of global self-attention. In contrast to previous works that concatenate the unimodal tokens along the sequence dimension, example approaches described herein align per-modality tokens by chaining them along the channels. Specifically, the tokens from one modality can be use...