Long-tail large scale face recognition by non-linear feature level domain adaption

1. A surveillance system with facial recognition, the surveillance system comprising:

one or more cameras;

a processor device and memory coupled to the processor device, the processing system programmed to:

receive a plurality of images from the one or more cameras;

extract, with a feature extractor utilizing a convolutional neural network (CNN) with an enlarged intra-class variance of long-tail classes, feature vectors from each of the plurality of images;

generate, with a feature generator, discriminative feature vectors for each of the feature vectors;

classify, with a fully connected classifier, an identity from the discriminative feature vectors; and

control an operation of a processor-based device to react in accordance with the identity.

2. The surveillance system as recited in claim 1 , further includes a communication system.

3. The surveillance system as recited in claim 2 , wherein the communication system connects to a remote server that includes a facial recognition network.

4. The surveillance system as recited in claim 1 , wherein the operation includes implementing a person containment procedure.

5. The surveillance system as recited in claim 1 , wherein the operation closes and locks doors and windows.

6. The surveillance system as recited in claim 1 , wherein the operation alerts authorities to an intruder.

7. The surveillance system as recited in claim 1 , wherein the one or more cameras is a body cam.

8. The surveillance system as recited in claim 1 , further programmed to train the feature extractor, the feature generator, and the fully connected classifier with an alternative bi-stage strategy.

9. The surveillance system as recited in claim 8 , wherein one stage of the alternative bi-stage strategy fixes the feature extractor and applies the feature generator to generate new transferred features that are more diverse and violate a decision boundary.

10. The surveillance system as recited in claim 8 , wherein one stage of the alternative bi-stage strategy fixes the fully connected classifier and updates the feature extractor and the feature generator.

11. The surveillance system as recited in claim 1 , wherein the feature extractor shares covariance matrices across all classes to transfer intra-class variance from regular classes to the long-tail classes.

12. The surveillance system as recited in claim 1 , wherein the feature generator optimizes a softmax loss by joint regularization of weights and features through a magnitude of an inner product of the weights and features.

13. The surveillance system as recited in claim 1 , wherein the feature extractor averages the feature vector with a flipped feature vector, the flipped feature vector being generated from a horizontally flipped frame from one of the plurality of images.

14. The surveillance system as recited in claim 1 , further programmed to control an operation of a processor-based machine to react in accordance with the identity.

15. The surveillance system as recited in claim 1 , wherein each of the plurality of images is selected from the group consisting of an image, a video, and a frame from the video.

16. A computer program product for a surveillance system with facial recognition, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:

receiving, by a processor device, a plurality of images from one or more cameras;

extracting, by the processor device with a feature extractor utilizing a convolutional neural network (CNN) with an enlarged intra-class variance of long-tail classes, feature vectors for each of the plurality of images;

generating, by the processor device with a feature generator, discriminative feature vectors for each of the feature vectors;

classifying, by the processor device utilizing a fully connected classifier, an identity from the discriminative feature vector; and

controlling an operation of a processor-based device to react in accordance with the identity.

17. A computer-implemented method for facial recognition in a surveillance system, the method comprising:

receiving, by a processor device, a plurality of images from one or more cameras;

extracting, by the processor device with a feature extractor utilizing a convolutional neural network (CNN) with an enlarged intra-class variance of long-tail classes, feature vectors for each of the plurality of images;

generating, by the processor device with a feature generator, discriminative feature vectors for each of the feature vectors;

classifying, by the processor device utilizing a fully connected classifier, an identity from the discriminative feature vector; and

controlling an operation of a processor-based device to react in accordance with the identity.

18. The computer-implemented method as recited in claim 17 , wherein controlling includes alerting authorities to an intruder.

19. The computer-implemented method as recited in claim 17 , wherein controlling includes implementing a person containment procedure.

20. The computer-implemented method as recited in claim 17 , wherein controlling includes closing and locking doors and windows.