Machine learning method, machine learning device, and machine learning program
a machine learning and program technology, applied in machine learning, image analysis, image enhancement, etc., can solve the problems of increasing the amount of calculation, limiting the configurable batch size, and sometimes exceeding the memory capacity of the gpu, so as to achieve high accuracy and increase the batch size
Active Publication Date: 2021-11-18
SHIMADZU CORP
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Benefits of technology
This approach enables increased batch sizes and accurate evaluation of learning models by generating full-size high-definition label images, overcoming memory constraints and pixel signal errors, thus enhancing the processing of large images and evaluation accuracy.
Problems solved by technology
In general, in deep learning for image data, the calculation amount becomes enormous.
However, even in this case, since it is difficult to process a large number of images at a time, processing is performed in which an input image is segmented into a predetermined number of images and a loss function is calculated for each group of the segmented images.
Increasing the batch size enhances the learning and the learning can be performed efficiently, but there is a limitation on a configurable batch size due to the restriction of a memory capacity of a GPU.
Further, when the size of the input image is extremely large, even if the batch size is set to a minimum value, the data size required for learning may sometimes exceed the memory capacity of the GPU and processing may not be executed.
In such a case, since the original cell observation image is very high in definition and wide in photographing range, the size of the image data is enormous, so that the temporary memory size used at the time of the learning becomes enormous.
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
[0030]Hereinafter, an example of a machine learning method, a machine learning device, and a machine learning program according to the present invention will be described with reference to the attached drawings.
[0031]FIG. 1 is a schematic block diagram of an example of a machine learning device for performing a machine learning method according to the present invention.
[0032]The machine learning device of this example includes, as functional blocks, an image reduction unit 11, a learning execution unit 12 including an FCN computation unit 13, an image enlargement unit 14, an error calculation unit 15, and a parameter update unit 16, and a learning control unit 17. Note that, this machine learning device may be embodied by making a computer system including a personal computer or a higher performance workstation, or a high-performance computer connected to such a computer via a communication line as a hardware resource and executing software installed on such a computer unit or a com...
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
A full-size training image is reduced by an image reduction unit (11) and input to an FCN (Fully Convolutional Neural Network) computation unit (13), and the FCN computation unit (13) performs calculation under a set filter coefficient and outputs a reduced label image. The reduced label image is enlarged to a full size by an image enlargement unit (14), the error calculation unit (15) calculates an error between the enlarged label image and a full-size ground truth based on a loss function, and the parameter update unit (16) updates a filter coefficient depending on the error. By repeating learning under the control of the learning control unit 17, it is possible to generate a learning model for performing optimal segmentation including an error generated at the time of image enlargement. Further, by including the image enlargement processing in the learning model, a full-size label image can be output, and the accuracy evaluation of the model can also be performed with high accuracy.
Description
TECHNICAL FIELD[0001]The present invention relates to a machine learning method and, a machine learning device, and more particularly, to a machine learning method and a machine learning device for generating a learning model for performing semantic segmentation on an image by supervised machine learning.BACKGROUND ART[0002]In recent years, in various fields of, e.g., automatic operation and medical image diagnosis, image processing techniques, such as, e.g., semantic segmentation using machine learning, in particular, deep learning, have been used. In the image processing using deep learning, a convolutional neural network (hereinafter abbreviated as “CNN”) has been widely used. The CNN usually includes a convolutional layer that extracts features of an image by convolution processing by a plurality of filters, a pooling layer that gives position invariance of local data by pooling processing that aggregates responses in a certain region, and a fully connected layer that connects i...
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 & AuthorityApplications(United States)
IPC IPC(8): G06T7/11G06N3/02G06N20/00
CPCG06T7/11G06N3/02G06T2207/20084G06T2207/20081G06N20/00G06T7/00G06N3/084G06N3/045
InventorTAKAHASHI, WATARUOSHIKAWA, SHOTA
OwnerSHIMADZU CORP