Systems and methods for two-stage objection detection

Patent metadata

Field	Value
Publication number	US-12555263-B2
Application number	US-202218063819-A
Country	US
Kind code	B2
Filing date	Dec 9, 2022
Priority date	Dec 9, 2022
Publication date	Feb 17, 2026
Grant date	Feb 17, 2026

Abstract

Example implementations include a method, apparatus and computer-readable medium for detecting an object in an image, by applying, on a first input image, a first object detector configured to generate a bounding box around the object in the first input image. The implementations further include identifying a first bounding box having a first bounding box size around the object in the first input image and determining that the first bounding box size is less than a threshold size. Additionally, the implementations further include generating a second input image by cropping an area of the original image corresponding to the first bounding box and identifying, by applying a second object detector on the second input image, a second bounding box around the object in the second input image. Additionally, the implementations further include performing an image analysis function on the object using information from the second bounding box.

First claim

What is claimed is:

1 . An apparatus for detecting an object in an image, comprising:
a memory; and a processor coupled with the memory and configured to:
generate a first input image for a first object detector by adjusting an original image size of an original image depicting the object to a first input image size, wherein the first object detector is a classifier configured to generate a bounding box around the object in the first input image;
identify, by applying the first object detector on the first input image, a first bounding box having a first bounding box size around the object in the first input image;
determine whether the first bounding box size is less than a threshold size;
generate, in response to determine that the first bounding box size is less than the threshold size, a second input image by enlarging and upsampling a cropped area of the original image corresponding to the first bounding box, wherein a second input image size of the second input image is greater than the first bounding box size and less than the first input image size;
identify, by applying a second object detector on the second input image, a second bounding box around the object in the second input image, wherein the second object detector is a second classifier having a second processing speed that is different than a first processing speed of the first object detector, and wherein a difference in processing speed between the second object detector and the first object detector is based on the second object detector having a different number of layers and/or neurons and processing different sized images relative to the first object detector; and
perform an image analysis function on the object using information from the second bounding box in the second input image.

2 . The apparatus of claim 1 , wherein adjusting the original image size comprises one or more of resizing, cropping, padding, or rotating.

3 . The apparatus of claim 1 , wherein the threshold size is a ratio of the first bounding box size relative to the first input image size.

4 . The apparatus of claim 3 , wherein the ratio is between 5%-20%.

5 . The apparatus of claim 1 , wherein the first input image size is less than the original image size and wherein the first input image size is larger than the second input image size.

6 . The apparatus of claim 1 , wherein adjusting the original image size comprises adjusting to meet first dimension requirements for the first object detector, and wherein to generate the second input image the processor is further configured to adjust the second input image size to meet second dimension requirements for the second object detector.

7 . The apparatus of claim 1 , wherein a second ratio of an object size of the object relative to the second bounding box is greater than a first ratio of the object size relative to the first bounding box.

8 . The apparatus of claim 1 , wherein the processor is further configured to:
perform, in response to determine that the first bounding box size is not less than the threshold size, the image analysis function on the object using information from the first bounding box in the first input image.

9 . The apparatus of claim 1 , wherein the image analysis function comprises one or more of keypoint detection, object classification, edge detection, segmentation, pose estimation, or noise filtering.

10 . The apparatus of claim 1 , wherein to identify, by applying the second object detector on the second input image, the second bounding box around the object in the second input image the processor is further configured to identify a set of keypoints associated with the object in the second image.

11 . The apparatus of claim 1 , wherein the object is a face of a person and wherein the image analysis function is a facial recognition algorithm, wherein the processor is further configured to:
compare, using the facial recognition algorithm, the face in the second bounding box with another face; and generate an alert indicative of a match in response to determining, based on the comparing, that the face and the another face match.

12 . The apparatus of claim 11 , wherein to identify, by applying the second object detector on the second input image, the second bounding box around the object in the second input image the processor is further configured to identify a set of keypoints associated with the object in the second image, and wherein comparing, using the facial recognition algorithm, the face in the second bounding box with the another face the processor is further configured to compare the set of keypoints with a known set of keypoints associated with the another face.

13 . The method of claim 1 , wherein the second object detector is a second classifier having a second processing speed that is faster than a first processing speed of the first object detector, and wherein a difference in processing speed between the second object detector and the first object detector is based on the second object detector having fewer layers and/or neurons and processing smaller images relative to the first object detector.

14 . A method for detecting an object in an image, comprising:
generating a first input image for a first object detector by adjusting an original image size of an original image depicting the object to a first input image size, wherein the first object detector is a classifier configured to generate a bounding box around the object in the first input image; identifying, by applying the first object detector on the first input image, a first bounding box having a first bounding box size around the object in the first input image; determining whether the first bounding box size is less than a threshold size; generating, in response to determining that the first bounding box size is less than the threshold size, a second input image by enlarging and upsampling a cropped area of the original image corresponding to the first bounding box, wherein a second input image size of the second input image is greater than the first bounding box size and less than the first input image size; identifying, by applying a second object detector on the second input image, a second bounding box around the object in the second input image, wherein the second object detector is a second classifier having a second processing speed that is different than a first processing speed of the first object detector, and wherein a difference in processing speed between the second object detector and the first object detector is based on the second object detector having a different number of layers and/or neurons and processing different sized images relative to the first object detector; and performing an image analysis function on the object using information from the second bounding box in the second input image.

15 . The method of claim 14 , wherein adjusting the original image size comprises one or more of resizing, cropping, padding, or rotating.

16 . The method of claim 14 , wherein the threshold size is a ratio of the first bounding box size relative to the first input image size.

17 . The method of claim 16 , wherein the ratio is between 5%-20%.

18 . The method of claim 14 , wherein the first input image size is less than the original image size and wherein the first input image size is larger than the second input image size.

19 . The method of claim 14 , wherein adjusting the original image size comprises adjusting to meet first dimension requirements for the first object detector, and w

Assignees

• Tyco Fire & Security Gmbh

Inventors

• Levy Ran
• Rozner Amit
• Falik Yohay

Classifications

• G06T3/02 Primary

  Affine transformations (for image registration G06T3/147; for image mosaicing G06T3/4038) · CPC title

• G06T7/13

  Edge detection · CPC title

• G06V40/167

  using comparisons between temporally consecutive images · CPC title

• G06V40/161

  Detection; Localisation; Normalisation · CPC title

• G06T7/73 Primary

  using feature-based methods · CPC title