Granted Patent B1

US 11,165,992 · App. 17/149,938 · Granted Nov 2, 2021

Inventors: Chin Leong Ong (Singapore, SG); Seng Khoon Teh (Singapore, SG)

Assignee: Dell Products L.P.

H04N7/157G06K9/00255G06K9/00275H04N7/147H04N7/152

Abstract

Claims (57)

1. A computer-implemented method comprising:

receiving a video frame from a video source, wherein the video frame includes faces of individuals engaged in a video conference;

generating, from the received video frame, a windowed image for each face in the video frame, the generating the windowed image comprising executing a crowd counting operation on the video frame to determine a location of each face in the video frame, the crowd counting operation using a trained crowd counting inference model;

generating a composite video frame including the windowed images; and

providing the composite video frame for use as video of a video-conferencing application.

2. The computer-implemented method of claim 1 , wherein the generating of the windowed images comprises:

executing a face detection operation using the location of each face as determined by the crowd counting operation, wherein the face detection operation extracts facial features of the faces located in the video frame.

3. The computer-implemented method of claim 2 , further comprising:

using the extracted features of each face to construct a facial image for each face of the individuals in the video frame.

4. The computer-implemented method of claim 3 , further comprising:

determining whether the video frame includes an obstructed facial image; and

executing a face recovery operation to recover facial features of the obstructed facial image, wherein the face recovery operation restores facial features of the obstructed facial image missing from the video frame to generate a recovered facial image.

5. The computer-implemented method of claim 4 , further comprising:

generating the windowed image for the obstructed facial image using the recovered facial image.

6. The computer-implemented method of claim 3 , further comprising:

scaling the facial images to generate scaled facial images for each of the individuals, wherein the scaled facial images have approximately the same dimensions.

7. The computer-implemented method of claim 5 , further comprising:

generating the windowed images using the scaled facial images.

8. A system comprising:

one or more information handling systems, wherein the one or more information handling systems include:

a processor;

a data bus coupled to the processor; and

a non-transitory, computer-readable storage medium embodying computer program code, the non-transitory, computer-readable storage medium being coupled to the data bus;

wherein the computer program code included in one or more of the information handling systems is executable by the processor of the information handling system so that the information handling system, alone or in combination with other information handling systems, executes operations comprising:

receiving a video frame from a video source, wherein the video frame includes faces of individuals engaged in a video conference;

generating, from the received video frame, a windowed image for each face in the video frame, the generating the windowed image comprising executing a crowd counting operation on the video frame to determine a location of each face in the video frame, the crowd counting operation using a trained crowd counting inference model;

generating a composite video frame including the windowed images; and

providing the composite video frame for use as video of a video-conferencing application.

9. The system of claim 8 , wherein the generating of the windowed images comprises:

executing a face detection operation using the location of each face as determined by the crowd counting operation, wherein the face detection operation extracts facial features of the faces located in the video frame.

10. The system of claim 9 , wherein the operations further comprise:

using the extracted features of each face to construct a facial image for each face of the individuals in the video frame.

11. The system of claim 10 , wherein the operations further comprise:

determining whether the video frame includes a face of an obstructed individual, wherein facial features of the obstructed individual are partially obscured; and

executing a face recovery operation to recover facial features of the obstructed individual, wherein the face recovery operation restores facial features of the obstructed individual missing from the video frame to generate a recovered facial image.

12. The system of claim 11 , wherein the operations further comprise:

generating the windowed image for the face of the obstructed individual using the recovered facial image.

13. The system of claim 10 , wherein the operations further comprise:

scaling the facial images to generate scaled facial images for each of the individuals, wherein the scaled facial images have approximately the same dimensions.

14. The system of claim 13 , wherein the operations further comprise:

generating the windowed images using the scaled facial images.

15. A non-transitory, computer-readable storage medium embodying computer program code, the computer program code comprising computer-executable instructions configured for executing operations comprising:

receiving a video frame from a video source, wherein the video frame includes faces of individuals engaged in a video conference;

generating, from the received video frame, a windowed image for each face in the video frame, the generating the windowed image comprising executing a crowd counting operation on the video frame to determine a location of each face in the video frame, the crowd counting operation using a trained crowd counting inference model;

generating a composite video frame including the windowed images; and

providing the composite video frame for use as video of a video-conferencing application.

16. The non-transitory, computer-readable storage medium of claim 15 , wherein the generating of the windowed images comprises:

executing a face detection operation using the location of each face as determined by the crowd counting operation, wherein the face detection operation extracts facial features of the faces located in the video frame.

17. The non-transitory, computer-readable storage medium of claim 16 , wherein the operations further comprise:

using the extracted features of each face to construct a facial image for each face of the individuals in the video frame.

18. The non-transitory, computer-readable storage medium of claim 17 , wherein the operations further comprise:

determining whether the video frame includes an obstructed facial image; and

executing a face recovery operation to recover facial features of the obstructed facial image, wherein the face recovery operation restores facial features of the obstructed facial image missing from the video frame to generate a recovered facial image.

19. The non-transitory, computer-readable storage medium of claim 18 , wherein the operations further comprise:

generating the windowed image for the obstructed facial image using the recovered facial image.

20. The non-transitory, computer-readable storage medium of claim 17 , wherein the operations further comprise:

scaling the facial images to generate scaled facial images for each of the individuals, wherein the scaled facial images have approximately the same dimensions.

Assignments (9)

Cited By (5)

US 12,223,766 US 12,299,845 US 12,381,929 US 12,444,228 US 12,477,078