US 12,033,391 B2
Systems and methods for detecting deep neural network inference quality using image/data manipulation without ground truth information
Gurjeet Singh, Castro Valley, CA (US); Apurbaa Mallik, Santa Clara, CA (US); Zafar Iqbal, Hayward, CA (US); Hitha Revalla, San Jose, CA (US); Steven Chao, San Francisco, CA (US); and Vijay Nagasamy, Fremont, CA (US)
Assigned to Ford Global Technologies, LLC, Dearborn, MI (US)
Filed by Ford Global Technologies, LLC, Dearborn, MI (US)
Filed on Dec. 10, 2021, as Appl. No. 17/547,495.
Prior Publication US 2023/0186637 A1, Jun. 15, 2023
Int. Cl. G06V 20/56 (2022.01); G06F 18/23 (2023.01); G06N 3/02 (2006.01); G06N 5/04 (2023.01)
CPC G06V 20/56 (2022.01) [G06F 18/23 (2023.01); G06N 3/02 (2013.01); G06N 5/04 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for inference quality determination of a deep neural network (DNN) comprising:
receiving an image frame from a source;
applying a normal inference DNN model to the image frame to produce a first inference with a first bounding box using a normal inference DNN model;
applying a deep inference DNN model to a plurality of filtered versions of the image frame to produce a plurality of deep inferences with a plurality of bounding boxes;
comparing the plurality of bounding boxes to identify a cluster condition of the plurality of bounding boxes; and
determining an inference quality of the image frame of the normal inference DNN model as a function of the cluster condition.