Defect sample generation method and device, storage medium and electronic device
By training a PGGAN network to generate defective block samples and fusing them with the original samples, the problem of low diversity of defective samples is solved, and the recognition accuracy and applicability of deep learning models are improved.
CN116645571BActive Publication Date: 2026-07-24GUOHUA ENERGY INVESTMENT
View PDF 2 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUOHUA ENERGY INVESTMENT
- Filing Date
- 2023-04-20
- Publication Date
- 2026-07-24
Smart Images
Figure CN116645571B_ABST
Abstract
The present disclosure relates to a kind of flaw sample generation method, device, storage medium and electronic equipment, the method includes: original defect sample is deformed and handled, to generate original flaw image, extract flaw area in original flaw image, generate original flaw data set, based on progressive growth adversarial generation PGGAN network training original flaw data set, to generate flaw block sample, with original non-flaw sample in area and predetermined image fusion of location, generate target flaw sample.Thereby, original flaw data set is trained and learned using PGGAN network, generate diverse flaw sample block, again, flaw sample block is fused with original non-flaw sample, improve the diversity of expansion sample.
Need to check novelty before this filing date? Find Prior Art