Scene and target identification method and device based on multi-task learning
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A multi-task learning and target recognition technology, applied in character and pattern recognition, instruments, biological neural network models, etc., can solve problems such as difficult and effective functions
Inactive Publication Date: 2017-06-13
珠海习悦信息技术有限公司
View PDF2 Cites 78 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
However, since the scene is related to multi-layer information such as the target, the background environment, and the spatial layout, the scene features of the same category have strong variability, and the scene features of different categories may have similar points overlapping each other, which brings great difficulties for accurate recognition. great challenge
At present, the overall scene recognition based on cutting-edge deep learning technology can only achieve a top-1 accuracy rate of about 50%, and it is difficult to play an effective role in practical applications.
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
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0084] Such as figure 1 As shown, a method for scene and target recognition based on multi-task learning, the method includes the following steps:
[0085] Step S1: Collect pictures containing different scenes and objects as image sample data;
[0086] Step S2: Manually label the image sample data to obtain the target category label and the scene category label;
[0087] Step S3: Construct a multi-layer convolutional neural network model and perform network initialization;
[0088] Step S4: Using the image sample data and the corresponding target category labels, pre-train the constructed model until convergence to obtain the target recognition model;
[0089] Step S5: Based on the multi-task learning technology, add network branches to a specific layer of the target recognition model, and initialize randomly to obtain a multi-task network;
[0090] Step S6: Retrain the multi-task network by using image sample data and corresponding scene category labels and target category...
Embodiment 2
[0093] Such as figure 1 As shown, a method for scene and target recognition based on multi-task learning, the method includes the following steps:
[0094] Step S1: Collect pictures containing different scenes and objects as image sample data; including the following steps:
[0095] Step S11: an image collection step, using cameras and network resources to collect image data of different scenes and objects;
[0096] Step S12: an image screening step, performing secondary screening on the image data, removing image data with unsatisfactory picture quality and screen content, and using the image data of the remaining images as image sample data. Remaining images ≥ 3000. Preferably, there are more than 20,000 remaining images.
[0097] Step S2: Manually label the image sample data to obtain the target category label and the scene category label; including the following steps:
[0098] Step S21: mark the target category, mark N_ob target category labels for each image, and sto...
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 relates to a scene and target identification method and device based on multi-task learning. The method comprises the steps that pictures containing different scenes and targets are collected as image sample data; the image sample data is subjected to manual label marking, and target class labels and scene class labels are obtained; a multi-layer convolutional neural network model is built, and network initialization is conducted; the image sample data and the corresponding target class labels are adopted for pre-training the built model till convergence, and a target identification model is obtained; based on a multi-task learning technology, network branches are added into a specific layer of the target identification model, random initialization is conducted, and a multi-task network is obtained; the image sample data and the corresponding scene class labels and target class labels are adopted for e-training the multi-task network till convergence, and a multi-task learning model is obtained; new image data is input to the multi-task learning model, and classification results of scene and target identification of images are obtained. Accordingly, the single task identification precision is improved.
Description
technical field [0001] The invention relates to the combination of the field of vision, image recognition and deep learning, in particular to a method and device for scene and target recognition based on multi-task learning. Background technique [0002] With the rise of deep learning, more and more technologies use deep learning to realize image recognition of pictures or video streams. Compared with traditional methods, deep learning avoids the complexity of manual parameter adjustment and manual feature selection. By building a deep network model, multi-layer analysis and abstract feature extraction are performed on data, which has high accuracy, high reliability, and high Adaptive features. Common image recognition applications cover action recognition, face recognition, object recognition, scene recognition, etc. Among them, object recognition and scene recognition, as the basis of image retrieval, image classification, scene understanding, and environment perception,...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to view more 
Login to view more