US 12,189,714 B2
System and method for improved few-shot object detection using a dynamic semantic network
Marios Savvides, Pittsburgh, PA (US); Chenchen Zhu, Pittsburgh, PA (US); Fangyi Chen, Pittsburgh, PA (US); Uzair Ahmed, Pittsburgh, PA (US); and Ran Tao, Pittsburgh, PA (US)
Assigned to Carnegie Mellon University, Pittsburgh, PA (US)
Appl. No. 18/266,744
Filed by CARNEGIE MELLON UNIVERSITY, Pittsburgh, PA (US)
PCT Filed Feb. 2, 2022, PCT No. PCT/US2022/014833
§ 371(c)(1), (2) Date Jun. 12, 2023,
PCT Pub. No. WO2022/173621, PCT Pub. Date Aug. 18, 2022.
Application 18/266,744 is a continuation of application No. 17/408,674, filed on Aug. 23, 2021, granted, now 12,026,226.
Claims priority of provisional application 63/068,871, filed on Aug. 21, 2020.
Claims priority of provisional application 63/147,782, filed on Feb. 10, 2021.
Prior Publication US 2024/0045925 A1, Feb. 8, 2024
Int. Cl. G06F 18/2136 (2023.01); G06N 3/04 (2023.01); G06N 5/02 (2023.01)
CPC G06F 18/2136 (2023.01) [G06N 3/04 (2013.01); G06N 5/02 (2013.01)] 22 Claims
OG exemplary drawing
 
1. A few-shot object detector comprising:
a visual network with trainable parameters producing an output based on an input of a visual feature; and
a dynamic semantic network which accepts as input a language feature representing a semantic representation of a class and outputs a class-specific parameter for the visual network;
wherein the dynamic semantic network includes a dynamic relation graph for building direct connections between base classes and novel classes using non-visual knowledge of the base classes.