Information processing apparatus and training method
An information processing apparatus, for training a classifier that classifies local regions of an object, includes a feature amount setting unit, a selection unit, and a training unit. The feature amount setting unit sets a feature amount to be used by the classifier. The selection unit selects a local region of the object based on a predetermined selection condition based on positions for obtaining the feature amount set by the feature amount setting unit. The training unit trains the classifier using the feature amount set by the feature amount setting unit and the local region selected by the selection unit.
1. An information processing apparatus for training a classifier that classifies local regions of an object, the information processing apparatus comprising:
a computer-readable memory; and
one or more processors that are coupled to the computer-readable memory and that are configured to cause the information processing apparatus to implement:
a feature amount setting unit configured to set a feature amount to be used by the classifier and positions for obtaining the feature amount,
a selection unit configured to select a local region of the object based on a rate of positions in the local region of the object to positions for obtaining the feature amount set by the feature amount setting unit, and
a training unit configured to train the classifier using the feature amount set by the feature amount setting unit and the local region selected by the selection unit.
2. The information processing apparatus according to claim 1 , further comprising a selection condition setting unit configured to set the rate.
3. The information processing apparatus according to claim 1 , wherein the feature amount is a result of a binary decision based on comparing a value obtained at a predetermined position and a value obtained at another predetermined position in the local region.
4. The information processing apparatus according to claim 1 , wherein the feature amount setting unit randomly sets the positions for obtaining the feature amount in the local region.
5. The information processing apparatus according to claim 1 , wherein the feature amount setting unit sets one of the positions for obtaining the feature amount in the local region at a center of the local region.
6. The information processing apparatus according to claim 1 , wherein the rate is a condition to be applied to a plurality of positions for obtaining the feature amount set by the feature amount setting unit.
7. The information processing apparatus according to claim 1 , further comprising a condition resetting unit configured to change, according to the local region selected by the selection unit, at least one of the feature amount and the rate,
wherein the selection unit reselects a local region of the object based on the feature amount or the selection condition change by the condition resetting unit, and
wherein the training unit trains the classifier using the local region reselected by the selection unit.
8. The information processing apparatus according to claim 1 , further comprising a determination unit configured to determine, based on the local region selected by the selection unit, whether a class of the object can be classified,
wherein the training unit trains the classifier using local regions from which local regions in the class determined unclassifiable by the determination unit are removed.
9. The information processing apparatus according to claim 1 , further comprising:
an image input unit configured to input a captured image of the object; and
a classification unit configured to classify, by using the classifier trained by the training unit, a class of the object in the image input by the image input unit.
10. The information processing apparatus according to claim 1 , further comprising a weight determination unit configured to determine, based on the local region selected by the selection unit, a weight to be assigned when each of the local regions is classified.
11. The information processing apparatus according to claim 10 , further comprising:
an image input unit configured to input a captured image of the object; and
a classification unit configured to discriminate, by using the classifier trained by the training unit, a class of the object in the image input by the image input unit, according to the weight determined by the weight determination unit.
12. The information processing apparatus according to claim 1 , further comprising:
an image input unit configured to input a captured image of the object;
a classification unit configured to classify, by using the classifier trained by the training unit, a class of the object in the image input by the image input unit; and
a condition resetting unit configured to change, according to a classification result by the classification unit, at least one of the feature amount and the rate,
wherein the selection unit reselects a local region of the object based on the feature amount or the selection condition reset by the condition resetting unit, and
wherein the training unit retrains the classifier using the local region reselected by the selection unit.
13. A method for an information processing apparatus for training a classifier that classifies local regions of an object, the method comprising:
setting a feature amount to be used by the classifier and positions for obtaining the feature amount;
selecting a local region of the object based on a rate of positions in the local region of the object to positions for obtaining the set feature amount; and
training, using one or more processors that are coupled to a computer-readable memory, the classifier using the set feature amount and the selected local region.
14. A non-transitory computer-readable storage medium storing a program causing a computer to perform the method according to claim 13 .