OBJECT RECOGNITION WITH REDUCED NEURAL NETWORK WEIGHT PRECISION

• US 20160086078A1
• Filed: 03/19/2015
• Published: 03/24/2016
• Est. Priority Date: 09/22/2014
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Ilia Ovsiannikov, Zhengping Ji
• Inventors: OVSIANNIKOV, Ilia, WANG, Yibing M., SHI, Lilong, JI, Zhengping
• Granted Patent: US 10,417,525 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06F 18/24137 Distances to cluster centroïds
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18057 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 30/19127 Extracting features by transforming the feature space, e.g. multidimensional scaling; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 30/19173 Classification techniques

Abstract

A client device configured with a neural network includes a processor, a memory, a user interface, a communications interface, a power supply and an input device, wherein the memory includes a trained neural network received from a server system that has trained and configured the neural network for the client device. A server system and a method of training a neural network are disclosed.

Claims

1. A client device configured with a neural network, the client device comprising:

a processor, a memory, a user interface, a communications interface, a power supply and an input device;

the memory comprising a trained neural network received from a server system that has trained and configured the neural network for the client device.

2. The client device as in claim 1, wherein the input device is configured to capture an image and to store image input data in the memory.

3. The client device as in claim 1, further comprising a multilayer perceptron (MLP) classifier configured to map image input data.

4. The client device as in claim 1, wherein the neural network comprises a convolutional neural network.

5. The client device as in claim 1, wherein the neural network is configured to generate a feature map.

6. The client device as in claim 5, wherein the feature map comprises a plurality of weight values derived from an input image.

7. The client device as in claim 5, wherein neural network is configured to perform a quantization operation on the feature map.

8. The client device as in claim 7, wherein the quantization operation comprises one of unitary quantization, adaptive quantization, uniform quantization, and supervised iterative quantization.

9. The client device as in claim 7, wherein the quantization performs back-propagation (BP) of image input data.

10. The client device as in claim 1, wherein the neural network is configured to perform object recognition.

11. The client device as in claim 1, comprising one of a smartphone, a tablet computer and a portable electronic device.

12. A system for providing object recognition with a client device, the system comprising:

a server system configured for training a neural network to perform object recognition and exporting the neural network to the client device.

13. The system as in claim 12, further comprising:

the client device comprising a processor, a memory, a user interface, a communications interface, a power supply and an input device; and

,

the memory comprising the trained neural network received from the server system.

14. The system as in claim 12, wherein the server system comprises a database of training images.

15. The system as in claim 12, wherein the neural network comprises a convolutional neural network.

16. The system as in claim 15, wherein the convolutional neural network comprises at least two layers.

17. The system as in claim 15, wherein the convolutional neural network comprises a convolutional layer and a subsampling layer.

18. The system as in claim 12, wherein the server system comprises a multilayer perceptron (MLP) classifier.

19. The system as in claim 12, wherein a learning technique for the neural network comprises one of a back-propagation (BP), Restricted Boltzmann Machine, and Auto Encoder Decode technique.

20. A server system, comprising:

an input device configured to receive a training image;

a neural network that includes at least two layer pairs, each layer pair comprising a convolutional layer and a subsampling layer;

a multilayer perceptron (MLP) classifier;

wherein the neural network is configured to perform quantization of interim weights in the convolutional layers, and

the neural network is also configured to generate in the subsampling layer an interim feature map in response to an input applied to the convolutional layer; and

the neural network configured to perform quantization of weights in the MLP classifier, and

the neural network is configured to generate in the MLP classifier a classification output in response to the feature map being applied to quantized weights MLP.