Back to results
US 11,386,637 B2Granted
Method and apparatus for detecting object
AssigneeSamsung Electronics Co., Ltd.
InventorsSeungin Park, Lin Ma, Weiming Li, Yang Liu, Hao Wang, Hyong Euk Lee
Life of this patent
Estimating the term…
Abstract
An object detection method includes setting a first window region and a second window region larger than the first window region that correspond to partial regions of different sizes in an input image, downsampling the second window region to generate a resized second window region, detecting a first object candidate from the first window region and a second object candidate from the resized second window region, and detecting an object included in the input image based on one or both of the first object candidate and the second object candidate.
Claims34
An object detection method comprising:
- setting a first window region and a second window region corresponding to partial regions of different sizes in an input image, wherein the second window region is larger than the first window region;
- downsampling the second window region to generate a resized second window region;
- detecting a first object candidate from the first window region and a second object candidate from the resized second window region;
- and
- detecting an object included in the input image based on one or both of the first object candidate and the second object candidate, the detecting of the object including removing an overlapping region from the first object candidate and the second object candidate based on a probability of the object being present in the first window region and/or the resized second window region,
- wherein the detecting of the first object candidate and the second object candidate comprises:
- in a case in which the second object candidate is not detected in a second image extracted from the resized second window region, detecting the first object candidate by applying a first image extracted from the first window region to a first neural network, and
- wherein the detecting of the object comprises:
- adjusting a location of the first window region based on a location of the second object candidate;
- detecting the first object candidate from the adjusted location of the first window region;
- and
- detecting the object included in the input image from the second object candidate and the first object candidate detected from the adjusted location.
9 more elements
An object detection apparatus comprising:
- a communication interface configured to obtain an input image;
- and
- a processor configured to:
- set a first window region and a second window region larger than the first window region that correspond to partial regions of different sizes in the input image;
- downsample the second window region to generate a resized second window region;
- detect a first object candidate from the first window region and a second object candidate from the resized second window region;
- and
- detect an object included in the input image based on one or both of the first object candidate and the second object candidate,
- wherein the processor is configured to:
- for the detecting of the object, remove an overlapping region from the first object candidate and the second object candidate based on a probability of the object being present in the first window region and/or the resized second window region;
- and
- in a case in which the second object candidate is not detected in a second image extracted from the resized second window region, detect the first object candidate by applying a first image extracted from the first window region to a first neural network, and
- wherein the detecting of the object comprises:
- adjusting a location of the first window region based on a location of the second object candidate;
- detecting the first object candidate from the adjusted location of the first window region;
- and
- detecting the object included in the input image from the second object candidate and the first object candidate detected from the adjusted location.
14 more elements
An object detection apparatus comprising:
- a communication interface configured to obtain an input image;
- and
- a processor configured to:
- set a first window region of a first size and a second window region of a second size larger than the first size corresponding to partial regions of the input image;
- generate a resized second window region from the second window region;
- determine whether a second object candidate is detected from a second image extracted from the resized second window region by applying the second image to a second neural network;
- determine whether to apply a first image extracted from the first window region to a first neural network to detect a first object candidate based on whether the second object candidate is detected from the second image;
- and
- detect an object included in the input image based on one or both of the first object candidate and the second object candidate,
- wherein the processor is configured to:
- for the detecting of the object, remove an overlapping region from the first object candidate and the second object candidate based on a probability of the object being present in the first window region and/or the resized second window region;
- and
- in a case in which the second object candidate is not detected in the second image, detect the first object candidate by applying the first image to the first neural network, and
- wherein the detecting of the object comprises:
- adjusting a location of the first window region based on a location of the second object candidate;
- detecting the first object candidate from the adjusted location of the first window region;
- and
- detecting the object included in the input image from the second object candidate and the first object candidate detected from the adjusted location.
15 more elements
Dependencies are inferred from the claim text.