Method for tracking target objects in a specific space, and device using the same

An obfuscation tracking device for tracking one or more objects in a specific space, comprising:

• at least one memory that stores instructions;
• and
• at least one processor configured to execute the instructions to perform or support another device to perform:
• (I) inputting original images of the specific space taken from at least one camera to an obfuscation network and instructing the obfuscation network to obfuscate the original images to generate obfuscated images such that the obfuscated images are not identifiable as the original images by a human but the obfuscated images are identifiable as the original images by a learning network;
• (II) inputting the obfuscated images into the learning network, and instructing the learning network to detect one or more obfuscated target objects, corresponding to one or more target objects to be tracked, in the obfuscated images, to thereby output information on the obfuscated target objects;
• and (III) tracking the obfuscated target objects in the specific space by referring to the information on the obfuscated target objects,
• wherein the obfuscation network has been trained to obfuscate the original images such that the obfuscated images are not identifiable as the original images by the human but the obfuscated images are identifiable as the original images by the learning network, and
• wherein a learning device has trained the obfuscation network by performing processes of (i) inputting training data into the obfuscation network and instructing the obfuscation network to generate obfuscated training data by obfuscating the training data, (ii) (ii-1) inputting the obfuscated training data into the learning network with one or more trained parameters and (ii-2) instructing the learning network to apply at least one network operation to the obfuscated training data by using the trained parameters, to thereby generate characteristic information corresponding to the obfuscated training data and (iii) training the obfuscation network such that at least one first error is minimized which is calculated by referring to (1) the characteristic information or task specific output generated by using the characteristic information and (2) its corresponding ground truth and (iii-2) at least one second error is maximized which is calculated by referring to the training data and the obfuscated training data.