US 12,395,338 B2
Electronic device for performing token pruning in frequency domain and method for operating the same
Hyun Kim, Seoul (KR); and Jong Ho Lee, Seoul (KR)
Assigned to Foundation for Research and Business, Seoul National University of Science and Technology, Seoul (KR)
Filed by Foundation for Research and Business, Seoul National University of Science and Technology, Seoul (KR)
Filed on Feb. 9, 2023, as Appl. No. 18/166,828.
Claims priority of application No. 10-2022-0184327 (KR), filed on Dec. 26, 2022.
Prior Publication US 2024/0214205 A1, Jun. 27, 2024
Int. Cl. G06T 5/00 (2024.01); G06T 5/10 (2006.01); G06T 7/11 (2017.01); G06V 10/26 (2022.01); H04L 9/32 (2006.01)
CPC H04L 9/3213 (2013.01) [G06T 5/10 (2013.01); G06T 7/11 (2017.01)] 12 Claims
OG exemplary drawing
 
1. An electronic device for performing token pruning in a frequency domain, the device comprising:
a processor,
wherein the processor is configured to,
divide an image frame into a plurality of patches,
convert tokens based on the plurality of patches from a spatial domain to a frequency domain through a fast Fourier transform (FFT)-based frequency domain conversion algorithm, the tokens being output from a predetermined transformer block among a plurality of transformer blocks, after performing patch embedding on the plurality of patches, and
convert the remaining tokens other than specific tokens from the frequency domain to the spatial domain, after pruning specific tokens from the tokens based on frequency information of the tokens converted into the frequency domain,
wherein the processor is configured to prune the specific tokens from the tokens by the number of the tokens selected by a user in the order of a token with the highest frequency to a token with the lowest frequency.