Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Object detection method, device and electronic apparatus

An object detection and object technology, which is applied in character and pattern recognition, instruments, computer components, etc., can solve the problem of difficult coverage of each object, and achieve the effects of small granularity, improved accuracy, and easy modeling

Inactive Publication Date: 2018-06-29
BEIJING SENSETIME TECH DEV CO LTD
View PDF2 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, in the above-mentioned method of obtaining candidate frames first and then performing image classification, because the existing method of generating candidate frames is difficult to ensure that every object in the image is covered, if an object is missed in the candidate frame generation step, Then the object can no longer be detected

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Object detection method, device and electronic apparatus
  • Object detection method, device and electronic apparatus
  • Object detection method, device and electronic apparatus

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0037] refer to figure 1 , shows a flow chart of steps of an object detection method according to Embodiment 1 of the present invention.

[0038] The object detection method of this embodiment includes the following steps:

[0039] Step S102: Obtain multiple superpixels in the image to be detected.

[0040] Wherein, the image to be detected is an image after superpixel segmentation. In this embodiment, a plurality of superpixels in the image are obtained from the image after superpixel segmentation. In the embodiments of the present invention, multiple means two or more.

[0041] In practical applications, any appropriate superpixel segmentation method can be used to obtain superpixels in a picture, including but not limited to superpixel segmentation methods based on graph theory, such as graph-based method, Ncut method, superpixel lattice method , methods based on entropy rate, etc.; or methods based on gradient descent, such as watershed method, MeanShift method, Quick-...

Embodiment 2

[0050] refer to figure 2 , shows a flow chart of steps of an object detection method according to Embodiment 2 of the present invention.

[0051] In this embodiment, an energy function for object detection is first trained, and then the energy function is used for image detection to determine the category and / or position of the object in the image. The object detection method of this embodiment includes the following steps:

[0052] Step S202: Obtain sample images for training.

[0053] Wherein, the sample image includes the information of the segmented superpixels.

[0054] Step S204: using the sample image to train an energy function.

[0055] In a feasible manner, the sample image can be used to train an energy function based on RCNN (Region based Convolutional Neural Network) and a set objective function until the energy function has a significant impact on the sample image. The annotations of the superpixels meet the set training termination conditions. The objectiv...

Embodiment 3

[0070] refer to image 3 , shows a flow chart of steps of an object detection method according to Embodiment 3 of the present invention.

[0071] In this embodiment, the energy function trained for object detection is further described. After the energy function is trained, the energy function is used for image detection to determine the category and / or location of the object in the image.

[0072] The object detection method of this embodiment includes the following steps:

[0073] Step S302: Obtain a sample image, perform superpixel segmentation on the sample image, and use the sample image after superpixel segmentation as a sample image for training.

[0074] In this embodiment, the superpixel segmentation result of each sample image used for training is denoted as P, where P={p 1 ,p 2 ,...,p N}, p i is the i-th superpixel, and N is the number of superpixels; if p i and p j connected together in space, then p i and p j belongs to a neighborhood system X, that is, ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The embodiments of the present invention provide an object detection method, an object detection device, and an electronic apparatus. The object detection method includes the following steps that: a plurality of super pixels in an image to be detected are acquired; the representation features of each super pixel and spatial context information between the plurality of super pixels are extracted; and the category and / or location of at least one object included in the image are / is determined according to the representation features and the spatial context information. With the object detection method and the object detection device of the embodiments of the invention adopted, object detection is carried out based on super pixels, and therefore, objects missed by a traditional candidate framegeneration algorithm can be naturally detected out, and the accuracy of object detection can be effectively improved.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of artificial intelligence, and in particular to an object detection method, device and electronic equipment. Background technique [0002] With the development of object detection technology, object detection provides solutions to many other high-level computer vision problems, and becomes the basis for solving computer vision problems, such as image search, face recognition, tracking and behavior recognition. [0003] At present, most of the work on object detection is to transform the object detection problem into the problem of image classification object. This process often first generates a candidate object region, that is, a candidate frame, and then independently classifies these candidate frames. For example, use a sliding window to sample multiple scales and positions to obtain approximately 100,000 candidate boxes per picture, or cluster or segment according to image feature...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/62
CPCG06F18/2415G06F18/214
Inventor 王昌宝闫俊杰
Owner BEIJING SENSETIME TECH DEV CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products