IP Library › Granted Patent US 12,731,279
Granted Patent B2
US 12,731,279 · App. 18/423,204 · Granted Sep 8, 2026

Segmentation-based generation of bounding shapes

Inventor: Michal Nir (Tel Aviv, IL)
Assignee: AUTOBRAINS TECHNOLOGIES LTD.
G06T7/60G06T7/12G06T7/194G06T7/50G06T7/70G06V10/26G06V10/82G06V20/58G06T2207/20081G06T2207/20084G06T2207/20132G06T2207/30256G06V20/588G06V2201/10
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,731,279
App. No.
18/423,204
Granted
Sep 8, 2026
Kind
B2
Abstract

A method that is computer implemented and for segmentation-based generation of bounding shapes, the method includes (i) obtaining a bounding shape that bounds an object captured in a sensed information unit; (ii) generating a cropped image in correspondence with the bounding shape; (iii) segmenting the cropped image to a cropped image background and a cropped image foreground; and (iv) generating an updated bounding shape in correspondence with the cropped image, using the segmentation, the updated bounding shape surrounding the cropped image foreground of the cropped image.

Claims (30)

1 . A method that is computer implemented and for segmentation-based generation of bounding shapes, the method comprises:

obtaining a bounding shape that bounds an object captured in a sensed information unit; wherein the obtaining comprises feeding the sensed information unit to an object detection neural network;

generating, by a vehicle processing circuit of the vehicle, a cropped image dedicated to the object in correspondence with the bounding shape;

segmenting the cropped image to a cropped image background and a cropped image foreground; and

generating, by the vehicle processing circuit, an updated bounding shape in correspondence with the cropped image, using the segmentation, the updated bounding shape surrounding the cropped image foreground of the cropped image; and

responding to the updated bounding shape by autonomously driving a vehicle, under a control of an autonomous control driving unit of the vehicle that comprises one or more processing circuits.

2 . The method according to claim 1 , further comprising processing the updated bounding shape to produce driving related content.

3 . The method according to claim 1 , wherein the responding further comprises determining a distance between the object and the vehicle, wherein the autonomously driving is responsive to the distance.

4 . The method according to claim 2 , wherein with the sensed information unit acquired using a sensor, the processing involves determining a distance between the sensor and the object captured in the sensed information unit.

5 . The method according to claim 4 , wherein the sensed information unit is at least one image, and wherein the determining of the distance comprises detecting a location of a horizon within the at least one image.

6 . The method according to claim 1 , wherein the sensed information unit captures another object, and the method further comprises obtaining another bounding shape that bounds the other object and generating another cropped image dedicated to the other object in correspondence with the other bounding shape.

7 . The method according to claim 1 , wherein the generated cropped image consists essentially of a corresponding object bounded by the bounding shape.

8 . The method according to claim 1 , wherein the generated cropped image consists essentially of (i) a corresponding object bounded by the bounding shape and (ii) a margin.

9 . The method according to claim 8 , wherein a dimension of the margin is up to 5 percent of a corresponding dimension of the bounding shape, the dimension is selected of a height and width.

10 . The method according to claim 8 , wherein at least one of a width and weight of the margin is up to 4 pixels.

11 . The method according to claim 1 , wherein the segmenting is agnostic to object metadata generated during a generation of the bounding shape.

12 . A non-transitory computer readable medium for segmentation-based generation of bounding shapes, the non-transitory computer readable medium stores instructions for:

obtaining a bounding shape that bounds an object captured in a sensed information unit; wherein the obtaining comprises feeding the sensed information unit to an object detection neural network;

generating, by a vehicle processing circuit of the vehicle, a cropped image dedicated to the object in correspondence with the bounding shape; the processor comprising one or more processing circuits;

segmenting the cropped image to a cropped image background and a cropped image foreground;

generating, by the vehicle processing circuit, an updated bounding shape in correspondence with the cropped image, using the segmentation, the updated bounding shape surrounding the cropped image foreground of the cropped image; and

responding to the updated bounding shape by autonomously driving the vehicle, under a control of an autonomous control driving unit of the vehicle that comprises one or more processing circuits.

13 . The non-transitory computer readable medium according to claim 12 , further comprising processing the updated bounding shape to produce driving related content.

14 . The non-transitory computer readable medium according to claim 12 , wherein the responding further comprises determining a distance between the object and the vehicle, wherein the autonomously driving is responsive to the distance.

15 . The non-transitory computer readable medium according to claim 13 , wherein with the sensed information unit acquired using a sensor, the processing involves determining a distance between the sensor and the object captured in the sensed information unit.

16 . The non-transitory computer readable medium according to claim 15 , wherein the sensed information unit is at least one image, and wherein the determining of the distance comprises detecting a location of a horizon within the at least one image.

17 . The non-transitory computer readable medium according to claim 12 , wherein the sensed information unit captures another object, and the non-transitory computer readable medium further stores instructions for obtaining another bounding shape that bounds the other object and generating another cropped image dedicated to the other object in correspondence with the other bounding shape.

18 . A vehicle processing circuit that is configured to: obtain a bounding shape that bounds an object captured in a sensed information unit wherein the obtaining comprises feeding the sensed information unit to an object detection neural network; generate a cropped image in correspondence with the bounding shape; segment the cropped image to a cropped image background and a cropped image foreground; and generate an updated bounding shape in correspondence with the cropped image, using the segmentation, the updated bounding shape surrounding the cropped image foreground of the cropped image; and respond to the updated bounding shape by autonomously driving the vehicle, under a control of an autonomous control driving unit of the vehicle that comprises one or more processing circuits.

19 . The vehicle processing circuit according to claim 18 , wherein the responding further comprises determining a distance between the object and the vehicle, wherein the autonomously driving is responsive to the distance.

20 . The vehicle processing circuit according to claim 18 , wherein the sensed information unit is at least one image, and wherein the vehicle processing unit if further configured to determine of the distance comprises detecting a location of a horizon within the at least one image.

Continuity (2)
Provisional Application 63481789 · Jan 26, 2023
Related Publication 20240257378A1 · Aug 1, 2024
References Cited (10)
US 20140056520A1 · Rodriguez Serrano · 2014 [cited by examiner]
US 20170297488A1 · Wang · 2017 [cited by examiner]
US 20180150716A1 · Farooqi · 2018 [cited by examiner]
US 20190130165A1 · Seshadri · 2019 [cited by examiner]
US 20190347501A1 · Kim · 2019 [cited by examiner]
US 20200175326A1 · Shen · 2020 [cited by examiner]
US 20200393845A1 · Wang · 2020 [cited by examiner]
US 20220351526A1 · Bar Zvi · 2022 [cited by examiner]
Wang et al, Detection and Classification of Moving Vehicle From Video Using Multiple Spatio-Temporal Features, 2019, IEEE Digital Object Identifier, pp. 1-13. (Year: 2019). [cited by examiner]
Chien et al, On Improving Bounding Box Regression Towards Accurate Object Detection Tracking, 2019 Int. Conference on Image and Vision Computing New Zealand, pp. 1-7. (Year: 2019). [cited by examiner]