Learning systems and methods
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.
1. A method comprising the acts:
applying a filtering operation to a set of image data depicting an object, captured by a camera, yielding filtered imagery;
transforming the filtered imagery from a pixel domain representation to a frequency domain representation, yielding frequency domain image data;
estimating scale and rotation parameters with which a digital watermark signal is depicted in the captured set of image data, by using correlation between said frequency domain image data and a frequency domain representation of a watermark orientation signal;
using said estimated scale and rotation parameters to identify locations at which to sample said filtered imagery to yield estimates of said digital watermark signal;
processing said sample estimates to extract digital watermark payload data, using an error-correcting decoding process;
associating an object-descriptive label with the captured imagery, said label comprising the extracted digital watermark payload data, or information looked-up in a database using said extracted digital watermark payload data;
repeating the aforesaid acts multiple times, to multiple sets of image data depicting multiple different objects, yielding a reference library in which multiple sets of image data are each associated with an object-descriptive label; and
using data from said library to train a deep learning network, comprising convolutional, max-pooling, and classification layers, to provide an output label for image data thereafter input to the deep learning network.
2. The method of claim 1 in which first and second sets of image data in said reference library are both associated with a common object-descriptive label, and a depiction of the object in the first set of image data differs from a depiction of the object in the second set of image data due to different viewing angle.
3. The method of claim 1 in which first and second sets of image data in said reference library are both associated with a common object-descriptive label, and a depiction of the object in the first set of image data differs from a depiction of the object in the second set of image data due to different focus.
4. The method of claim 1 in which first and second sets of image data in said reference library are both associated with a common object-descriptive label, and a depiction of the object in the first set of image data differs from a depiction of the object in the second set of image data due to different illumination.
5. The method of claim 1 that further includes capturing a first set of image data depicting a first object under strongly colored illumination, and capturing a second set of image data depicting a second object under strongly colored illumination.
6. A system comprising:
a camera, one or more processors and one or more memories, the one or more memories containing instructions configuring the system to perform acts including:
applying a filtering operation to a set of image data depicting an object, captured by the camera, yielding filtered imagery;
transforming the filtered imagery from a pixel domain representation to a frequency domain representation, yielding frequency domain image data;
estimating scale and rotation parameters with which a digital watermark signal is depicted in the captured set of image data, by using correlation between said frequency domain image data and a frequency domain representation of a watermark orientation signal;
using said estimated scale and rotation parameters to identify locations at which to sample said filtered imagery to yield estimates of said digital watermark signal;
processing said sample estimates to extract digital watermark payload data, using an error-correcting decoding process;
associating an object-descriptive label with the captured imagery, said label comprising the extracted digital watermark payload data, or information looked-up in a database using said extracted digital watermark payload data;
repeating the aforesaid acts multiple times, to multiple sets of image data depicting multiple different objects, yielding a reference library in which multiple sets of image data are each associated with an object-descriptive label; and
using data from said library to train a deep learning network, comprising convolutional, pooling, and classification layers, to provide an output label for image data thereafter input to the deep learning network.
7. A non-transitory computer readable medium containing instructions that configure a computer system to perform acts including:
applying a filtering operation to a set of image data depicting an object, captured by a camera, yielding filtered imagery;
transforming the filtered imagery from a pixel domain representation to a frequency domain representation, yielding frequency domain image data;
estimating scale and rotation parameters with which a digital watermark signal is depicted in the captured set of image data, by using correlation between said frequency domain image data and a frequency domain representation of a watermark orientation signal;
using said estimated scale and rotation parameters to identify locations at which to sample said filtered imagery to yield estimates of said digital watermark signal;
processing said sample estimates to extract digital watermark payload data, using an error-correcting decoding process;
associating an object-descriptive label with the captured imagery, said label comprising the extracted digital watermark payload data, or information looked-up in a database using said extracted digital watermark payload data;
repeating the aforesaid acts multiple times, to multiple sets of image data depicting multiple different objects, yielding a reference library in which multiple sets of image data are each associated with an object-descriptive label; and
using data from said library to train a deep learning network, comprising convolutional, pooling, and classification layers, to provide an output label for image data thereafter input to the deep learning network.