Granted Patent B2

US 12,646,318 · App. 18/173,779 · Granted Jun 2, 2026

Inventors: Andrei Boiarov (Sofia, BG); Ilya Shimchik (Istanbul, TR); Nikita Firsakov (Istanbul, TR); Pavlo Bredikhin (Kharkov, UA); Sergey Ulasen (Singapore, SG); Serg Bell (Singapore, SG); Stanislav Protasov (Singapore, SG); Nikolay Dobrovolskiy (Sofia, BG)

Assignees: Constructor Technology AG; Constructor Education and Research Genossenschaft

G06V20/41G06V10/774G06V10/945G06V20/48G06V20/49G06V20/70G06V20/44G06V2201/09

Abstract

Claims (33)

1 . A method for performing brand detection in a video, the method comprising:

receiving the video with a video splitter;

splitting the video with the video splitter to obtain a plurality of video frames;

providing the plurality of video frames to a brand detector for performing open set detection on each input video frame from the plurality of video frames, wherein the open set detection comprises:

proposing, by a localization module one or more bounding boxes on the input video frames on regions of a video frame that potentially include brand media,

cropping by a cropping module the one or more bounding boxes from the input video frames to obtain cropped bounding boxes,

providing the cropped bounding boxes to a classification module to obtain embedding vectors corresponding to each of the cropped bounding boxes, and

comparing, with a comparator module, the embedding vectors of the cropped bounding boxes with embedding vectors of one or more brand reference images provided by a user for computing instances of brand detection in each video frame of the plurality of video frames;

segmenting the brand media within the cropped bounding boxes by determining an exact region in which a brand logo is occupied within the cropped bounding boxes by a semantic segmentation model, wherein the exact region is only an area of the brand logo within the cropped bounding boxes; and

computing, by a brand appearance computing unit, one or more parameters associated with per-brand, per-appearance statistics of the brand logo in the video using the comparing.

2 . The method of claim 1 , wherein the brand media and the one or more brand reference images include brand logos, brand taglines, and brand ambassador images.

3 . The method of claim 1 , further comprising training the classification module in an open set approach using self-supervised learning (Supervised Contrastive learning) and few-shot learning.

4 . The method of claim 1 , further comprising resolving a scene understanding task by the semantic segmentation model.

5 . The method of claim 1 , further comprising detecting whether the one or more brand reference images appear in the video at a crucial moment by a video action recognition module.

6 . The method of claim 5 , further comprising identifying whether the one or more brand reference images appear in an area of a screen where a user's attention is focused by the video action recognition module.

7 . The method of claim 1 , further comprising allowing the user to label new brand reference images in the video frames for retraining the classification module.

8 . The method of claim 1 , further comprising allowing the user to provide new brand reference images for retraining the classification module.

9 . A system for performing brand detection in a video, the system comprising:

a video splitter to receive the video for performing the brand detection thereon, the video splitter configured to split the video to obtain a plurality of video frames;

a brand detector for performing an open set detection on each input video frame from the plurality of video frames, wherein the brand detector comprises:

a localization module to propose one or more bounding boxes on the input video frames on regions of a video frame that potentially include a brand media,

a cropping module to crop the one or more bounding boxes from the input video frames to obtain cropped bounding boxes,

a classification module to receive the cropped bounding boxes and obtaining embedding vectors corresponding to each of the cropped bounding boxes, and

a comparator module to compare the embedding vectors of the cropped bounding boxes with embedding vectors of one or more brand reference images provided by a user for computing instances of brand detection in each video frame of the plurality of video frames;

a semantic segmentation model to determine an exact area occupied by a brand logo within the cropped bounding boxes, wherein the exact area is only an area of the brand logo within the cropped bounding boxes; and

a brand appearance computing unit to compute one or more parameters associated with per-brand, per-appearance statistics of the brand logo in the video based on the comparing by the comparator module.

10 . The system of claim 9 , wherein the brand media and the one or more brand reference images include brand logos, brand taglines, and brand ambassador images.

11 . The system of claim 9 , wherein the classification module is trained in an open set approach using self-supervised learning (Supervised Contrastive learning) and few-shot learning.

12 . The system of claim 11 , wherein the semantic segmentation model is configured to perform a scene understanding task.

13 . The system of claim 9 , further comprising a video action recognition module to detect whether the one or more brand reference images appear in the video at a crucial moment.

14 . The system of claim 13 , wherein the video action recognition module is further configured to identify whether the one or more brand reference images appear in an area of a screen where a user's attention is focused.

15 . The system of claim 9 , further comprising a user interface to allow the user to label new brand reference images in the video frames for retraining the classification module.

16 . The system of claim 15 , wherein the user interface is further configured to allow the user to provide the new brand reference images for retraining the classification module.

Assignments (3)

Continuity (1)

References Cited (20)