Combined light and heavy models for image filtering

1. A method for demographic determination using image recognition, comprising:

analyzing an image with a pre-trained lightweight neural network model, where the lightweight neural network model generates a confidence value;

comparing the confidence value to a threshold value to determine if the pre-trained lightweight neural network model is sufficiently accurate;

analyzing the image with a pre-trained heavyweight neural network model for the confidence value below the threshold value, wherein the pre-trained heavyweight neural network model has above about one million trainable parameters and the pre-trained lightweight neural network model has a number of trainable parameters below one tenth the heavyweight model; and

displaying demographic data to a user on a user interface, wherein the user modifies store inventory based on the demographic data.

2. The method as recited in claim 1 , further comprising calculating the demographic data from a plurality of images.

3. The method as recited in claim 1 , wherein the pre-trained lightweight neural network model and the pre-trained heavyweight neural network model are each a convolutional neural network.

4. The method as recited in claim 3 , wherein the pre-trained lightweight neural network model is between 25% and 33% faster than the pre-trained heavyweight neural network model.

5. The method as recited in claim 1 , wherein the threshold value is set equal to the average of confidence values generated for the pre-trained heavyweight neural network model using a training set of images.

6. The method as recited in claim 1 , wherein the threshold value is 0.75.

7. A system for demographic determination using image recognition, comprising:

a memory, wherein a pre-trained lightweight neural network model and a pre-trained heavyweight neural network model are stored in the memory;

one or more processors configured to execute the pre-trained lightweight neural network model and the pre-trained heavyweight neural network model, wherein the pre-trained heavyweight neural network model has above about one million trainable parameters and the pre-trained lightweight neural network model has a number of trainable parameters below one tenth the heavyweight model, and compare a confidence value generated by the pre-trained lightweight neural network model to a threshold value, wherein the one or more processors are configured to execute the pre-trained heavyweight neural network model if the pre-trained lightweight neural network model is sufficiently accurate; and

a user interface configured to display demographic data to a user, wherein the user modifies store inventory based on the demographic data.

8. The system as recited in claim 7 , wherein the pre-trained lightweight neural network model has less than 1/10 th the trainable parameters than the pre-trained heavyweight neural network model.

9. The system as recited in claim 7 , wherein the pre-trained lightweight neural network model and the pre-trained heavyweight neural network model are each a convolutional neural network.

10. The system as recited in claim 7 , wherein the pre-trained lightweight neural network model is selected from the group consisting of Cifar10_quick, NECLA, MobileNet and Squeezenet.

11. The system as recited in claim 7 , wherein the pre-trained lightweight neural network model is selected from the group consisting of NECLA, AlexNet, ResNet, and VGGnet.

12. The system as recited in claim 7 , wherein the pre-trained lightweight neural network model is between 25% and 33% faster than the pre-trained heavyweight neural network model.

13. The system as recited in claim 7 , wherein the threshold value is 0.75.

14. The system as recited in claim 7 , wherein the threshold value is set equal to the average of confidence values generated for the pre-trained heavyweight neural network model using a training set of images.

15. A non-transitory computer readable storage medium comprising a computer readable program for demographic determination using image recognition, wherein the computer readable program when executed on a computer causes the computer to perform:

analyzing an image with a pre-trained lightweight neural network model, where the lightweight neural network model generates a confidence value;

comparing the confidence value to a threshold value to determine if the pre-trained lightweight neural network model is sufficiently accurate;

analyzing the image with a pre-trained heavyweight neural network model for the confidence value below the threshold value, wherein the pre-trained heavyweight neural network model has above about one million trainable parameters and the pre-trained lightweight neural network model has a number of trainable parameters below one tenth the heavyweight model; and

displaying demographic data to a user on a user interface, wherein the user modifies store inventory based on the demographic data.

16. The computer readable storage medium comprising a computer readable program, as recited in claim 15 , further comprising calculating the demographic data from a plurality of images.

17. The computer readable storage medium comprising a computer readable program, as recited in claim 15 , wherein the pre-trained lightweight neural network model and the pre-trained heavyweight neural network model are each a convolutional neural network.

18. The computer readable storage medium comprising a computer readable program, as recited in claim 17 , wherein the pre-trained lightweight neural network model is between 25% and 33% faster than the pre-trained heavyweight neural network model.

19. The computer readable storage medium comprising a computer readable program, as recited in claim 15 , wherein the threshold value is set equal to the average of confidence values generated for the pre-trained heavyweight neural network model using a training set of images.

20. The computer readable storage medium comprising a computer readable program, as recited in claim 15 , wherein the threshold value is 0.75.