Neural network tracing arrangements

Patent metadata

Field	Value
Publication number	US-2023325960-A1
Application number	US-202318181502-A
Country	US
Kind code	A1
Filing date	Mar 9, 2023
Priority date	Aug 2, 2013
Publication date	Oct 12, 2023
Grant date	—

Abstract

A sequence of images depicting an object is captured, e.g., by a camera at a point-of-sale terminal in a retail store. The object is identified, such as by a barcode or watermark that is detected from one or more of the images. Once the object's identity is known, such information is used in training a classifier (e.g., a machine learning system) to recognize the object from others of the captured images, including images that may be degraded by blur, inferior lighting, etc. In another arrangement, such degraded images are processed to identify feature points useful in fingerprint-based identification of the object. Feature points extracted from such degraded imagery aid in fingerprint-based recognition of objects under real life circumstances, as contrasted with feature points extracted from pristine imagery (e.g., digital files containing label artwork for such objects). A great variety of other features and arrangements—some involving designing classifiers so as to combat classifier copying—are also detailed.

First claim

1 - 30 . (canceled)

31 . A method employing a trained neural network, comprising the acts:
querying a first neural network for information, the first neural network comprising plural layers including neuron and pooling layers, the first neural network being characterized by multiple weighting parameters learned in a previously-conducted training operation employing first training data; receiving output data from the first neural network in response to said querying; detecting presence of a telltale sign in the received output data; and from said presence of the telltale sign in the received output data:
(a) establishing that the received output data was produced by said first neural network; or
(b) establishing that the first training data included said telltale sign, said first training data being of a type that does not normally include the telltale sign.

32 . The method of claim 31 that includes, from said detected presence of the telltale sign in the received output data, establishing that the received output data was produced by said first neural network.

33 . The method of claim 31 that includes, from said detected presence of the telltale sign in the received output data, establishing that the first training data included said telltale sign, said first training data being of a type that does not normally include the telltale sign.

34 . The method of claim 31 in which the received output data comprises image data, and the telltale sign comprises a predetermined marking.

35 . The method of claim 34 in which the predetermined marking comprises a text symbol.

36 . The method of claim 31 in which the telltale sign comprises a sequence of symbols.

37 . The method of claim 31 in which said querying act comprises querying the first neural network to determine an optimum stimulus, and the method further includes employing the first neural network to generate said optimum stimulus in response to said querying.

38 . The method of claim 31 in which the telltale sign is independent, within a feature space, from the first training images.

39 . The method of claim 31 wherein presence of the telltale sign in the received output data is not evident to a human but is detectable upon computer analysis.

40 . The method of claim 39 that includes conducting said computer analysis for detecting presence of the telltale sign in the received output data.

41 . The method of claim 40 in which said computer analysis comprises a statistical analysis of the received output data.

42 . The method of claim 40 in which said computer analysis comprises a correlation operation applied to the received output data.

43 . The method of claim 39 in which the received output data comprises an output image, and the telltale sign exploits limits of human visual acuity so that presence of the telltale sign in the received output data is not evident to a human observer of the output image.

44 . The method of claim 39 in which the received output data comprises an output image, and the telltale sign comprises a predetermined marking having a spatial frequency above 50 cycles per degree.

45 . The method of claim 31 in which the telltale sign encodes payload data employing error-correcting coding.

46 . The method of claim 31 in which the received output data comprises an output image and the telltale sign comprises a predetermined marking, wherein detecting presence of said predetermined marking in the received output data is conducted not in a pixel (spatial image) domain, but in a transform domain.

47 . The method of claim 31 in which the received output data comprises an output image and the telltale sign comprises a predetermined marking, wherein detecting presence of said predetermined marking in the received output data is conducted not in a pixel (spatial image) domain, but in a spatial frequency domain.

48 . The method of claim 31 in which the received output data comprises an output image and the telltale sign comprises a predetermined marking including a constellation of signal pulses at different spatial frequencies.

Assignees

• Digimarc Corp

Inventors

• Rodriguez Tony F
• Alattar Osama M
• Brunk Hugh L
• Meyer Joel R
• Conwell William Y
• Kamath Ajith M

Classifications

• G06T1/0021 Primary

  Image watermarking · CPC title

• G06F21/16

  Program or content traceability, e.g. by watermarking · CPC title

• G06V10/245

  by locating a pattern; Special marks for positioning · CPC title

• G06V20/20

  in augmented reality scenes · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title