Target detection method and apparatus

1. A method of detecting a target, the method comprising:

generating an image pyramid based on an image on which a detection is to be performed;

classifying a plurality of candidate areas in the image pyramid using a cascade neural network; and

determining a target area corresponding to a target included in the image based on the plurality of candidate areas,

wherein the cascade neural network comprises a plurality of neural networks, and at least one neural network among the plurality of neural networks comprises a plurality of parallel sub-neural networks.

2. The method of claim 1 , wherein the classifying comprises:

classifying a plurality of areas in the image using a first neural network; and

classifying the plurality of areas into a plurality of target candidate areas and a plurality of non-target candidate areas using a second neural network including the plurality of parallel sub-neural networks, and

wherein the plurality of neural networks comprises the first neural network and the second neural network.

3. The method of claim 2 , wherein the determining comprises:

normalizing positions and sizes of the plurality of target candidate areas based on layer images of the image pyramid comprising the plurality of target candidate areas and a difference in size and position between the layer images; and

acquiring the target area by merging a plurality of normalized target candidate areas.

4. The method of claim 1 , wherein each of the plurality of parallel sub-neural networks corresponds to a different target attribute.

5. The method of claim 4 , wherein, in response to the target included in the image being a human face, the target attribute includes any one or any combination of two or more of a front face posture, a side face posture, a front face or a side face by rotation, a skin color, a light condition, an occlusion, and a clarity.

6. The method of claim 1 , wherein the plurality of neural networks comprise a convolutional neural network and a Boltzmann network.

7. A non-transitory computer readable storage medium storing instructions that when actuated by a processor, cause the processor to perform the method of claim 1 .

8. The method of claim 1 , further comprising:

actuating a camera to capture the image on which the detection is to be performed; and,

actuating a processor to generate the image pyramid, classify the plurality of candidate areas; and determine the target area.

9. The method of claim 1 , further comprising:

recognizing a human face in the image comprising the target.

10. An apparatus for detecting a target, the apparatus comprising:

an image acquirer configured to generate an image pyramid based on an image on which a detection is to be performed;

a candidate area classifier configured to classify a plurality of candidate areas in the image pyramid using a cascade neural network; and

a target area determiner configured to determine a target area corresponding to a target included in the image based on the plurality of candidate areas,

wherein the cascade neural network comprises a plurality of neural networks, and at least one neural network among the plurality of neural networks comprises a plurality of parallel sub-neural networks.

11. The apparatus of claim 10 , wherein the candidate area classifier comprises:

a first classifier configured to classify a plurality of areas in the image using a first neural network; and

a second classifier configured to classify the plurality of areas into a plurality of target candidate areas and a plurality of non-target candidate areas using a second neural network comprising the plurality of parallel sub-neural networks, and

wherein the plurality of neural networks comprises the first neural network and the second neural network.

12. The apparatus of claim 11 , wherein the target area classifier is configured to normalize positions and sizes of the plurality of target candidate areas based on layer images of the image pyramid including the plurality of target candidate areas and a difference in size and position between the layer images, and acquire the target area by merging a plurality of normalized target candidate areas.

13. The apparatus of claim 10 , wherein each of the plurality of parallel sub-neural networks corresponds to a different target attribute.

14. The apparatus of claim 13 , wherein, in response to the target included in the image being a human face, the target attribute includes any one or any combination of two or more of a front face posture, a side face posture, a front face or a side face by rotation, a skin color, a light condition, an occlusion, and a clarity.

15. The apparatus of claim 10 , wherein the plurality of neural networks includes a convolutional neural network and a Boltzmann network.