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:
providing plural images as training images to a learning system;
decoding machine-readable identifiers from said plural images;
providing information based on the decoded machine-readable identifiers to the learning system for use in training the learning system with the training images; and
training the learning system with information obtained about the training images from the decoded machine-readable identifiers wherein the machine-readable identifiers comprise digital watermarks, the learning system comprises a neural network, wherein the neural network is incorporated into a digital watermark detector to improve digital watermark detection in images as measured by a signal confidence metric, and the method further comprises the acts of:
decoding digital watermarks from the plural images, yielding payload data;
using the payload data to assess the plural images from which the payload data is decoded; and
employing the payload data in training the neural network.
2. The method of claim 1 , in which said training includes, using backpropagation, adjusting weights in a multi-layer neural network that includes plural convolutional layers and pooling layers.
3. The method of claim 1 , comprising: training the neural network to incorporate a digital watermark in the neural network for use in determining whether the neural network has been copied.
4. The method of claim 1 , comprising: training the neural network to respond to images in which digital watermarks are embedded by providing an output that indicates whether the neural network has been used to create the output.
5. The method of claim 1 , wherein the signal confidence metric comprises a degree of correspondence between an extracted digital watermark and an expected digital watermark.
6. A non-transitory computer-readable storage medium having computer-executable instructions stored thereon, which when executed by one or more processors, cause the one or more processors to perform operations comprising:
decoding machine-readable identifiers from plural images provided as training images to a learning system;
providing information based on the decoded machine-readable identifiers to the learning system for use in training the learning system with the training images; and
training the learning system with information obtained about the training images from the decoded machine-readable identifiers;
wherein the machine-readable identifiers comprise digital watermarks, the learning system comprises a neural network, wherein the neural network is incorporated into a digital watermark detector to improve digital watermark detection in images as measured by a signal confidence metric, and the operations further comprise:
decoding digital watermarks from the plural images, yielding payload data;
using the payload data to assess the plural images from which the payload data is decoded; and
employing the payload data in training the neural network.
7. The non-transitory computer-readable storage medium of claim 6 , wherein the operations comprise: training the neural network to incorporate a digital watermark in the neural network for use in determining whether the neural network has been copied.
8. The non-transitory computer-readable storage medium of claim 6 , wherein the operations comprise: training the neural network to respond to images in which digital watermarks are embedded by providing an output that indicates whether the neural network has been used to create the output.
9. The non-transitory computer-readable storage medium of claim 6 , wherein the signal confidence metric comprises a degree of correspondence between an extracted digital watermark and an expected digital watermark.
10. A method comprising the acts:
for each of plural training images:
decoding digital watermark payload data from the training images;
providing information based on the payload data to a learning system, that comprises a neural network, for use in training the learning system with the training images; and
wherein the neural network comprises a multi-layer neural network,
training a classifier using said images and said payload data, said training including using the payload data to assess the plural images from which the payload data is decoded, and adjusting weights in the multi-layer neural network that includes plural convolutional layers and pooling layers, using backpropagation, the adjusted weights causing the network to respond to an image with an indication of an attribute associated with a digital watermark, wherein the neural network is incorporated into a digital watermark detector to improve digital watermark detection in images as measured by a signal confidence metric.
11. The method of claim 10 , in which said backpropagation employs gradient descent.
12. A learning system comprising:
one or more processors and memory configured with training images, a neural network, wherein the neural network comprises a multi-layer neural network, and computer-executable instructions to perform operations of:
decoding digital watermark payload data from the training images;
providing information based on the payload data to the multi-layer neural network for use in training the multi-layer neural network with the training images; and
training the multi-layer neural network using said images and said payload data, said training including using the payload data to assess the plural images from which the payload data is decoded and adjusting weights in the multi-layer neural network that includes plural convolutional layers and pooling layers, using backpropagation, the adjusted weights causing the network to respond to an image with an indication of an attribute associated with a digital watermark, wherein the neural network is incorporated into a digital watermark detector to improve digital watermark detection in images as measured by a signal confidence metric.
13. The learning system of claim 12 , wherein the computer-executable instructions are configured to: train the neural network to incorporate a digital watermark in the neural network for use in determining whether the neural network has been copied.
14. The learning system of claim 12 , wherein the computer-executable instructions are configured to: train the neural network to respond to images in which digital watermarks are embedded by providing an output that indicates whether the neural network has been used to create the output.
15. The learning system of claim 12 , wherein herein the signal confidence metric comprises a degree of correspondence between an extracted digital watermark and an expected digital watermark.
16. A method comprising the acts:
at a first time, decoding digital watermark payload data encoded in plural training images, and determining attribute metadata for plural of said training images using digital watermark payload data decoded from the plural training images;
at a second time after the first time, capturing images, providing the captured images to a trained neural network, and receiving from the trained neural network an indication of attribute metadata for said captured images;
wherein said neural network had been trained, between said first and second times, using the digital watermark payload data and the attribute metadata determined, at the first time;
the neural network comprising a multi-layer neural network that has been trained using said images and said digital watermark payload data, said training including using the digital watermark payload data to assess the plural images from which the payload data is decoded and adjusting weights in the multi-layer neural network that includes plural convolutional layers and pooling layers, using backpropagation, the adjusted weights causing the network to respond to an image with an indication of an attribute associated with a digital watermark, wherein the neural network is incorporated into a digital watermark detector to improve digital watermark detection in images as measured by a signal confidence metric.
17. The method of claim 16 , wherein said neural network had been trained, between the first and second times, using degraded depictions of the plural training images.