Unlock instant, AI-driven research and patent intelligence for your innovation.

Cascaded convolutional neural network training method, device and system and cascaded convolutional neural network based image detection method, device and system

A convolutional neural network and training method technology, applied in the field of image data processing, can solve problems such as unsatisfactory overall performance of multi-level and multi-layer neural networks

Active Publication Date: 2016-11-09
SHENZHEN SENSETIME TECH CO LTD
View PDF6 Cites 42 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in traditional cascaded networks, different levels of neural networks are usually trained separately, which can only achieve local optimization of each level of neural networks, and the overall performance of multi-level multi-layer neural networks is not ideal.

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
  • Cascaded convolutional neural network training method, device and system and cascaded convolutional neural network based image detection method, device and system
  • Cascaded convolutional neural network training method, device and system and cascaded convolutional neural network based image detection method, device and system
  • Cascaded convolutional neural network training method, device and system and cascaded convolutional neural network based image detection method, device and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0050] In order to achieve global optimal training of multi-level neural networks, this embodiment discloses a cascaded convolutional neural network training method, please refer to figure 1 , is the flow chart of the cascaded convolutional neural network training method, the method includes the following steps:

[0051] Step S110, acquiring image data of at least a local area of ​​the image to be learned. In a specific embodiment, a sliding window selection method may be used to select at least a partial area of ​​the image to be learned as the learning area of ​​the image. In an optional embodiment, each learning area can be circled by, for example, a square bounding box, and according to the degree of coincidence with the true value of the bounding box of the object area in the image, mark whether the learning area contains an object with detection, so as to facilitate Neural network learning training. In this embodiment, each learning area can be adjusted to a preset sta...

Embodiment 2

[0065] This embodiment discloses an image detection method based on a cascaded convolutional neural network, please refer to image 3 , is the flow chart of the image detection method based on the cascaded convolutional neural network, and the detection method includes the following steps:

[0066] Step S210, training a cascaded convolutional neural network model. In this embodiment, the neural network can be trained according to the cascaded convolutional neural network training method disclosed in Embodiment 1 to obtain a cascaded convolutional neural network model. It should be noted that, in this embodiment, step S10 is performed when training the neural network, and this step may not be performed after the training of the neural network is completed.

[0067] Step S220, acquiring image data of the image to be detected. In a specific embodiment, the image data can be preprocessed in advance to obtain the image data of the preprocessed image to be detected. Optionally, a ...

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 invention discloses a cascaded convolutional neural network training method, device and system and a cascaded convolutional neural network based image detection method, device and system. The training method comprises the steps of processing image data of at least a local region of an image to be learnt into image data of N types of input regions which are different in size, wherein N is an integer which is greater than or equal to 2; taking the image data of the N types of input regions as input of each-grade convolutional neural network in the N-grade cascaded convolutional neural network respectively, and carrying out training on each-grade convolutional neural network; and correlating at least one training result outputted by each-grade convolutional neural network respectively, transmitting the correlated training results back to each-grade convolutional neural network so as to adjust parameters of each-grade neural network. According to the invention, the parameters of each-grade neural network can be adjusted when the training results are broadcast to each-grade convolutional neural network, so that the cascaded convolutional neural network is enabled to achieve global optimization of the neural network parameters in training.

Description

technical field [0001] The invention relates to the field of image data processing, in particular to a cascaded convolutional neural network training and image detection method, device and system. Background technique [0002] Object detection is to accurately detect the position of all certain types of objects for the input picture, which plays an important role in the field of computer vision and pattern recognition. [0003] The traditional object detection method based on convolutional neural network first selects a series of regions to be detected with different positions and sizes on the picture, and then directly inputs the region into a convolutional neural network to obtain the classification result. By properly designing the structure of the convolutional neural network, the computer can directly learn the hidden features in the picture, avoiding artificial design features, and can be more widely used in the detection of various types of objects. However, since th...

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): G06K9/66G06N3/08G06T7/00
CPCG06N3/084G06N3/086G06T2207/20081G06T2207/20084G06V30/194
Inventor 秦红伟闫俊杰
Owner SHENZHEN SENSETIME TECH CO LTD