Instance segmentation method based on key points
A technology of key points and segmentation algorithm, which is applied in the field of workpiece recognition and positioning and attitude estimation system, and the field of workpiece recognition and positioning and attitude estimation system based on deep learning, to achieve the effect of strengthening contour segmentation, good segmentation, and improving segmentation accuracy
Pending Publication Date: 2020-08-07
NINGBO INST OF MATERIALS TECH & ENG CHINESE ACADEMY OF SCI +1
View PDF6 Cites 5 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0008] For the instance segmentation of complex scene images, the main purpose of the present invention is to add the concept of object key points on the basis of existing segmentation algorithms, and propose a more accurate instance segmentation algorithm based on key points to overcome the deficiencies of the prior art
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0047]In view of the deficiencies in the existing technology, the inventor of this case was able to propose the technical solution of the present invention after long-term research and a lot of practice, which mainly selects the basic Mask-RCNN algorithm and uses the mask as the training input, and calculates the key points of the object through the mask , and then use both masks and keypoints as input for training. The inventors of this case also discussed the effectiveness of different key point calculation methods to enhance the generalization and effectiveness of the algorithm. The technical solution, its implementation process and principle will be further explained as follows.
[0048] An aspect of the embodiments of the present invention provides a method for instance segmentation based on key points, which includes:
[0049] At least use the manually labeled mask map as the training input of the instance segmentation algorithm to obtain the object mask;
[0050] Usin...
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
Login to View More Abstract
The invention discloses an instance segmentation method based on key points, and the method comprises the steps: obtaining an object mask through employing at least a manually marked mask image as thetraining input of an instance segmentation algorithm; calculating to obtain a mask loss value through a mask loss function by utilizing the object mask and the manually marked mask image; adopting akey point algorithm to calculate key points of the manually marked mask image as the truth value image to obtain a truth value key point image; obtaining a prediction mask graph by using an instance segmentation algorithm, and calculating a key point loss value by using the prediction mask graph and the true value key point graph based on a set key point loss function; and optimizing network parameters in an instance segmentation algorithm based on the mask loss value and the key point loss value. According to the embodiment segmentation method, a key point algorithm is added on the basis of aMask-RCNN basic framework, a better effect can be achieved on segmentation of detail parts, and the segmentation precision of an object contour in a complex scene is improved.
Description
technical field [0001] The present invention relates to a system and method for workpiece identification and positioning and attitude estimation, in particular to a system and method for workpiece identification and positioning and attitude estimation based on deep learning, belonging to the field of target identification and detection. Background technique [0002] The instance segmentation algorithm is an image understanding method that detects and segments each object in the image individually. Different from the other two image understanding methods - semantic segmentation and target detection, semantic segmentation is to segment different objects without distinguishing between different individuals of the same object, that is, the same object gets a segmentation frame; target detection is Only detect one or several objects that need to be detected. Therefore, the accuracy and difficulty of instance segmentation are greater than other methods. [0003] In the developme...
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
Login to View More IPC IPC(8): G06K9/32G06K9/34G06K9/62
CPCG06V10/26G06V10/25G06F18/24Y02P90/30
Inventor 陈帅印李现周昊宇肖江剑
Owner NINGBO INST OF MATERIALS TECH & ENG CHINESE ACADEMY OF SCI