# US 2022/0019774 A1 · High resolution face matching

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0019774 A1
- Document ID: US20220019774A1
- Status: Published application
- Kind: A1
- Application number: 17/305,814
- Filed: 2021-07-14
- Published: 2022-01-20
- Earliest claimed: 2020-07-14
- Assignees: Corsight.Ai
- Applicants: Corsight.Ai
- Inventors: Ran Vardimon; Yarden Yaniv; Matan Noga; Gadi Piran
- Web page: https://patentgrep.com/patent/US20220019774A1

## CPC

- G06V 40/172: {Classification, e.g. identification}
- G06V 40/173: {face re-identification, e.g. recognising unknown faces across different face tracks}
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06T 19/00: Manipulating three-dimensional [3D] models or images for computer graphics
- G06V 10/82: using neural networks
- G06V 40/165: {using facial parts and geometric relationships}
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

Systems, and method and computer readable media that store instructions for high-resolution face recognition.

## Claims

### Claim 1

A method for high resolution face matching, the method comprises:
obtaining a visual representation of a first face and a visual representation of a second face; wherein the visual representations of the first and second faces are substantially aligned and are substantially of a same size;
processing the visual representation of the first face by first part of a convolutional neural network (CNN) to provide a first three-dimensional (3D) output; wherein the first 3D output comprises multiple first feature vectors, wherein different first feature vectors are associated with different areas of the visual representation of the first face;
processing the visual representation of the second face by the first part of the CNN to provide a second 3D output; wherein the second 3D output comprises multiple second feature vectors, wherein different second feature vectors are associated with different areas of the visual representation of the second face;
determining similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face; and
responding to the generation of the parts relevancy indications.

### Claim 2 (depends on 1)

The method according to claim 1 wherein the first part of the CNN consists essentially of the input layer of the CNN and only some of the intermediate layers of the CNN.

### Claim 3 (depends on 1)

The method according to claim 1 wherein the first part of the CNN consists essentially of the input layer of the CNN and less than half of the intermediate layers of the CNN.

### Claim 4 (depends on 1)

The method according to claim 1 wherein the first 3D output and the second 3D output are outputted from a single intermediate layer of the first part of the CNN.

### Claim 5 (depends on 1)

The method according to claim 1 wherein the first 3D output and the second 3D output are outputted a plurality of intermediate layers of the first part of the CNN.

### Claim 6 (depends on 1)

The method according to claim 1 wherein the first 3D output and the second 3D output comprises information regarding image properties that do not amount to object detection conclusions.

### Claim 7 (depends on 1)

The method according to claim 1 wherein the first 3D output and the second 3D output differ from object detection outputs

### Claim 8 (depends on 1)

The method according to claim 1 wherein different second feature vectors are associated with different second key points, the second key points are mapped to different areas of the visual representation of the second face; and wherein different first feature vectors are associated with different first key points, the first key points are mapped to different areas of the visual representation of the first face.

### Claim 9 (depends on 8)

The method according to claim 8 wherein the determining of the similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face comprises calculating similarities between a first feature vector associated with a first key point and between a second feature vector associated with a second key point that is located at a same location as the first key point.

### Claim 10 (depends on 8)

The method according to claim 8 wherein the determining of the similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face comprises calculating similarities between a first feature vector associated with a first key point and between second feature vectors associated with second key points, wherein the second key points are located within a search window that includes a second key point that is located at a same location as the first key point.

### Claim 11

A computerized system comprising a processor that is configured to:
obtain a visual representation of a first face and a visual representation of a second face; wherein the visual representations of the first and second faces are substantially aligned and are substantially of a same size;
process the visual representation of the first face by first part of a convolutional neural network (CNN) to provide a first three-dimensional (3D) output; wherein the first 3D output comprises multiple first feature vectors, wherein different first feature vectors are associated with different areas of the visual representation of the first face;
process the visual representation of the second face by the first part of the CNN to provide a second 3D output; wherein the second 3D output comprises multiple second feature vectors, wherein different second feature vectors are associated with different areas of the visual representation of the second face;
determine similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face; and
respond to the generation of the parts relevancy indications.

### Claim 12

A non-transitory computer readable medium for high resolution face matching, the non-transitory computer readable medium stores instructions for:
obtaining a visual representation of a first face and a visual representation of a second face; wherein the visual representations of the first and second faces are substantially aligned and are substantially of a same size;
processing the visual representation of the first face by first part of a convolutional neural network (CNN) to provide a first three-dimensional (3D) output; wherein the first 3D output comprises multiple first feature vectors, wherein different first feature vectors are associated with different areas of the visual representation of the first face;
processing the visual representation of the second face by the first part of the CNN to provide a second 3D output; wherein the second 3D output comprises multiple second feature vectors, wherein different second feature vectors are associated with different areas of the visual representation of the second face;
determining similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face; and
responding to the generation of the parts relevancy indications.

### Claim 13 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first part of the CNN consists essentially of the input layer of the CNN and only some of the intermediate layers of the CNN.

### Claim 14 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first part of the CNN consists essentially of the input layer of the CNN and less than half of the intermediate layers of the CNN.

### Claim 15 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first 3D output and the second 3D output are outputted from a single intermediate layer of the first part of the CNN.

### Claim 16 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first 3D output and the second 3D output are outputted a plurality of intermediate layers of the first part of the CNN.

### Claim 17 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first 3D output and the second 3D output comprises information regarding image properties that do not amount to object detection conclusions.

### Claim 18 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein the first 3D output and the second 3D output differ from object detection outputs

### Claim 19 (depends on 12)

The non-transitory computer readable medium according to claim 12 wherein different second feature vectors are associated with different second key points, the second key points are mapped to different areas of the visual representation of the second face; and wherein different first feature vectors are associated with different first key points, the first key points are mapped to different areas of the visual representation of the first face.

### Claim 20 (depends on 8)

The non-transitory computer readable medium according to claim 8 wherein the determining of the similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face comprises calculating similarities between a first feature vector associated with a first key point and between a second feature vector associated with a second key point that is located at a same location as the first key point.

### Claim 21 (depends on 8)

The non-transitory computer readable medium according to claim 8 wherein the determining of the similarities between the different areas of the visual representation of the second face and the different areas of the visual representation of the second face comprises calculating similarities between a first feature vector associated with a first key point and between second feature vectors associated with second key points, wherein the second key points are located within a search window that includes a second key point that is located at a same location as the first key point.
