Information processing device
An image recognition device includes: an image processing device that acquires a feature amount from an image; and an identification device that determines whether a prescribed identification object is present in the image, and identifies the identification object. The identification device includes a BNN that has learned the identification object in advance, and performs identification processing by performing a binary calculation with the BNN on the feature amount acquired by the image processing device. Then, the identification device selects a portion effective for identification from among high-dimensional feature amounts output by the image processing device to reduce the dimensions used in identification processing, and copies low-dimensional feature amounts output by the image processing device to increase dimensions.
1. An information processing device comprising:
a feature amount acquiring means configured to acquire a feature amount of identification object data;
a selection means configured to select a feature amount of a portion specified in advance from the feature amount acquired by the feature amount acquiring means;
an identification means including a binary neural network that has: (i) an input layer composed of multiple input units, (ii) an intermediate layer, and (iii) an output layer configured to learn an identification object based on multiple-value weighting in the binary neural network;
an input means configured to input each feature amount of the portion selected by the selection means into all of the input units of the binary neural network; and
an output means configured to output an identification result identified by the identification means based on the feature amount of the portion input by the input means.
2. The information processing device according to claim 1 , wherein in the identification means, the learning of the identification object is conducted by binarized weighting.
3. The information processing device according to claim 1 , wherein
the selection means selects a feature amount of a portion specified by an identification algorithm, in advance, from the feature amount acquired by the feature amount acquiring means, and
the identification algorithm is Real AdaBoost (RAdB).
4. The information processing device according to claim 3 , wherein
the selection means selects a feature amount of a portion, in which identification accuracy by the identification means becomes high, specified by the identification algorithm in advance, from the feature amount acquired by the feature amount acquiring means.
5. The information processing device according to claim 3 , wherein
the feature amount acquiring means acquires a feature amount based on distribution of co-occurrence of a luminance gradient extracted by a feature amount extraction means from the image data which is identification object data, and
the selection means selects a feature amount of a portion in which extraction processing or an extraction circuit configuration by the feature amount extraction means is simplified, specified by the identification algorithm in advance, from the feature amount acquired by the feature amount acquiring means.
6. The information processing device according to claim 1 , wherein the binary neural network includes: (a) an adder configured to binarize and add the feature amount of the portion, and (b) a counter configured to calculate an output of the adder.
7. An information processing device comprising:
a feature amount acquiring means configured to acquire a feature amount of identification object data by extracting a luminance gradient distribution of the identification object data;
a selection means configured to select a feature amount of a portion specified in advance from the feature amount acquired by the feature amount acquiring means;
a duplication means configured to duplicate the feature amount of the portion selected by the selection means;
an identification means configured to have learned an identification object based on multiple-valued weighting in a binary neural network;
an input means configured to input the feature amount of the portion selected by the selection means and the feature amount duplicated by the duplication means into the binary neural network; and
an output means configured to output an identification result of being identified by the identification means based on the feature amount of the portion input by the input means.
8. An information processing device comprising:
a processor programmed to:
acquire a feature amount of identification object data;
select a feature amount of a portion specified in advance from the acquired feature amount;
learn an identification object based on multiple-value weighting in a binary neural network, the binary neural network including: (i) an input layer composed of multiple input units, (ii) an intermediate layer, and (iii) an output layer;
input each feature amount of the selected portion into all of the input units of the binary neural network; and
output an identification result of the binary neural network based on the inputted feature amount of the selected portion.
9. The information processing device according to claim 8 , wherein the learning of the identification object is conducted via binarized weighting.
10. The information processing device according to claim 8 , wherein
the feature amount of the portion is selected via an identification algorithm, in advance, from the acquired feature amount, and
the identification algorithm is Real AdaBoost (RAdB).
11. The information processing device according to claim 10 , wherein
when an identification accuracy of the identification algorithm is greater than a predetermined amount for a current feature amount of the portion from the acquired feature amount, the identification algorithm selects the current feature amount as the selected feature amount of the portion.
12. The information processing device according to claim 10 , wherein
the feature amount is acquired based on a distribution of co-occurrence of a luminance gradient extracted from the image data that is in the identification object data, and
the feature amount of portion is selected based on extraction processing or an extraction circuit configuration from the acquired feature amount.
13. The information processing device according to claim 8 , wherein the binary neural network includes: (a) an adder configured to binarize and add the feature amount of the portion, and (b) a counter configured to calculate an output of the adder.