Classifier trained with data of different granularity
View Patent ↗A computer implemented method, computer system and computer program product are provided for data classification. According to the method, an original data is received by one or more processing units. A classification of the original data with a trained classifier is determined by one or more processing units, wherein the classifier is trained by a labeled data of different granularity, and the labeled data include at least one defect object.
1. A computer-implemented method, comprising:
receiving, by one or more hardware processors, an original data; and
determining, by the one or more hardware processors, a classification of the original data with a trained classifier,
wherein the classifier is trained by labeled data, which includes at least one defect object that is represented at a plurality of different size scales, in an order of granularity, performed over multiple epochs, with full images taking a longer stage of epochs than other size scales to improve stability.
2. The method of claim 1 , wherein the classifier is a convolutional neural networks (CNN).
3. The method of claim 1 , wherein the original data and the labeled data are images, and the at least one defect object is selected from: a defect part of a semiconductor device, a lesion within a biological tissue, a defect part of a Micro Electro Mechanical System (MEMS), and a defect area within a metal, and a stain in a textile printing material or image.
4. The method of claim 1 , wherein the labeled data of different size scales are labeled images with differing numbers of pixels.
5. The method of claim 1 , wherein the labeled data of different size scales are patches randomly cropped from labeled original data according to different scales.
6. The method of claim 1 , wherein training in the order of the granularity is from a fine granularity to a coarse granularity.
7. The method of claim 1 , wherein the labeled data of different size scales includes data of at least five levels of granularity according to their fineness.
8. The method of claim 1 , wherein the labeled data of different size scales has been preprocessed by at least one of: rotating; resizing; or masking.
9. The method of claim 1 , wherein the at least one defect object is a gastric ulcer.
10. A computer system, comprising:
a processor;
a computer-readable memory coupled to the processor, the memory comprising instructions that when executed by the processor perform actions of:
receiving an original data;
determining a classification of the original data with a trained classifier,
wherein the classifier is trained by labeled data, which includes at least one defect object that is represented at a plurality of different size scales, in an order of granularity, performed over multiple epochs, with full images taking a longer stage of epochs than other size scales to improve stability.
11. The system of claim 10 , wherein the original data and the labeled data are images, and the at least one defect object is selected from: a defect part of a semiconductor device, a lesion within a biological tissue, a defect part of a Micro Electro Mechanical System (MEMS), a defect area within a metal, and a stain in a textile printing material or image.
12. The system of claim 10 , wherein the labeled data of different size scales are labeled images with differing numbers of pixels.
13. The system of claim 10 , wherein the labeled data of different size scales are patches randomly cropped from a labeled original data according to different scales.
14. The system of claim 10 , wherein the classifier is trained with the labeled data of different size scales in order from a fine granularity to a coarse granularity.
15. A computer program product, comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
receiving an original data;
determining a classification of the original data with a trained classifier,
wherein the classifier is trained by labeled data, which includes at least one defect object that is represented at a plurality of different size scales, in an order of granularity, performed over multiple epochs, with full images taking a longer stage of epochs than other size scales to improve stability, and the labeled data include at least one defect object.
16. The product of claim 15 , wherein the original data and the labeled data are images, and the at least one defect object is selected from: a defect part of a semiconductor device, a lesion within a biological tissue, a defect part of a Micro Electro Mechanical System (MEMS), a defect area within a metal, and a stain in a textile printing material or image.
17. The product of claim 15 , wherein the labeled data of different size scales are labeled images with differing numbers of pixels.
18. The product of claim 15 , wherein the labeled data of different size scales are patches randomly cropped from labeled original data according to different scales.
19. The product of claim 15 , wherein the classifier is trained with the labeled data of different size scales in order from a fine granularity to a coarse granularity.