Method for performing target detection based on full convolutional network of deformable parts
A convolutional network and target detection technology, which is applied in the field of target detection based on a fully convolutional network based on deformable parts, can solve the problems of low resolution of feature maps and low target detection accuracy, and achieve the effect of improving positioning accuracy.
Inactive Publication Date: 2018-03-16
SHENZHEN WEITESHI TECH
View PDF3 Cites 8 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] Aiming at the problems of low feature map resolution and low target detection accuracy, the present invention utilizes a fully convolutional network, which is shared in the calculation of the entire image, provides task-related feature maps, uses a special RoI set to infer the region position, and finally Prediction for simple summarization, classification and localization for object detection
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
[0025] specific implementation plan
[0026] It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present invention will be further described in detail below in conjunction with the drawings and specific embodiments.
[0027] figure 1 It is a flowchart of a method for object detection based on a fully convolutional network of deformable parts in the present invention. It mainly includes fully convolutional feature extraction (1); RoI pooling based on deformable parts (2); classification and location prediction of deformable parts (3). Fully convolutional network for deformable parts, regional part representations, alignment by optimizing their positions, improving classification and localization prediction, part-based representations are more invariant to local transformations, part structure provides important information about object geometry .
[0028] The ...
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 provides a method for performing target detection based on a full convolutional network of deformable parts. The method mainly comprises full convolutional feature extraction, a deformable part-based RoI pool, and classification and locating prediction of the deformable parts. By taking the full convolutional network (FCN) as a backbone network structure, regional suggestions R are divided into a few parts by adopting the deformable part-based RoI pool; optimal matching shapes of local objects are located; alignment is performed in a corresponding position of an image; and prediction is performed by classifying and locating the regional suggestions by adopting two branches. By utilizing the FCN, sharing is realized in calculation of the whole image; feature mapping related totasks is provided; a regional position is inferred by adopting a special RoI set; and final prediction is simply summarized to obtain classification and locating for performing target detection. Thematching degree between the classification and detection tasks is ensured; the resolution of a feature graph is ensured; and the target detection accuracy is improved.
Description
technical field [0001] The invention relates to the field of object detection, in particular to a method for object detection based on a fully convolutional network of deformable parts. Background technique [0002] In recent years, deep learning based on deep convolutional networks has been widely used in some visual fields, mainly for image segmentation, semantic segmentation, and object detection. Target detection has become one of the key technologies in the field of artificial intelligence, and has a wide range of applications in visual navigation, intelligent transportation, video retrieval and compression, 3D reconstruction, security monitoring, and medical treatment. It is widely used in civil large scene fields. The commonly used region proposal network at this stage focuses on the region of interest in the image, classifies and locates the region, this method still has many deficiencies, it is prone to mismatch between classification and detection tasks, and reduc...
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 Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06K9/46G06N3/04
CPCG06V10/40G06N3/045G06F18/24
Inventor 夏春秋
Owner SHENZHEN WEITESHI TECH