Improved Training of CAE-Based Defect Detectors Using Structural Noise
Appearances of products are important to companies as they reflect the quality of their manufacture to customers. Nowadays, visual inspection is conducted by human inspectors. This research attempts to automate this process using Convolutional AutoEncoders (CAE). Our models were trained using images...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-12-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/11/24/12062 |
| _version_ | 1827674249616162816 |
|---|---|
| author | Reina Murakami Valentin Grave Osamu Fukuda Hiroshi Okumura Nobuhiko Yamaguchi |
| author_facet | Reina Murakami Valentin Grave Osamu Fukuda Hiroshi Okumura Nobuhiko Yamaguchi |
| author_sort | Reina Murakami |
| collection | DOAJ |
| description | Appearances of products are important to companies as they reflect the quality of their manufacture to customers. Nowadays, visual inspection is conducted by human inspectors. This research attempts to automate this process using Convolutional AutoEncoders (CAE). Our models were trained using images of non-defective parts. Previous research on autoencoders has reported that the accuracy of image regeneration can be improved by adding noise to the training dataset, but no extensive analyse of the noise factor has been done. Therefore, our method compares the effects of two different noise patterns on the models efficiency: Gaussian noise and noise made of a known structure. The test datasets were comprised of “defective” parts. Over the experiments, it has mostly been observed that the precision of the CAE sharpened when using noisy data during the training phases. The best results were obtained with structural noise, made of defined shapes randomly corrupting training data. Furthermore, the models were able to process test data that had slightly different positions and rotations compared to the ones found in the training dataset. However, shortcomings appeared when “regular” spots (in the training data) and “defective” spots (in the test data) partially, or totally, overlapped. |
| first_indexed | 2024-03-10T04:36:53Z |
| format | Article |
| id | doaj.art-a9b653d20dc24083a1446970420368e5 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-10T04:36:53Z |
| publishDate | 2021-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-a9b653d20dc24083a1446970420368e52023-11-23T03:42:39ZengMDPI AGApplied Sciences2076-34172021-12-0111241206210.3390/app112412062Improved Training of CAE-Based Defect Detectors Using Structural NoiseReina Murakami0Valentin Grave1Osamu Fukuda2Hiroshi Okumura3Nobuhiko Yamaguchi4Faculty of Science and Engineering, Saga University, Saga 840-8502, JapanFaculty of Science and Engineering, Saga University, Saga 840-8502, JapanFaculty of Science and Engineering, Saga University, Saga 840-8502, JapanFaculty of Science and Engineering, Saga University, Saga 840-8502, JapanFaculty of Science and Engineering, Saga University, Saga 840-8502, JapanAppearances of products are important to companies as they reflect the quality of their manufacture to customers. Nowadays, visual inspection is conducted by human inspectors. This research attempts to automate this process using Convolutional AutoEncoders (CAE). Our models were trained using images of non-defective parts. Previous research on autoencoders has reported that the accuracy of image regeneration can be improved by adding noise to the training dataset, but no extensive analyse of the noise factor has been done. Therefore, our method compares the effects of two different noise patterns on the models efficiency: Gaussian noise and noise made of a known structure. The test datasets were comprised of “defective” parts. Over the experiments, it has mostly been observed that the precision of the CAE sharpened when using noisy data during the training phases. The best results were obtained with structural noise, made of defined shapes randomly corrupting training data. Furthermore, the models were able to process test data that had slightly different positions and rotations compared to the ones found in the training dataset. However, shortcomings appeared when “regular” spots (in the training data) and “defective” spots (in the test data) partially, or totally, overlapped.https://www.mdpi.com/2076-3417/11/24/12062artificial intelligenceimage processingfeature detectionautoencoderconvolutional autoencoder |
| spellingShingle | Reina Murakami Valentin Grave Osamu Fukuda Hiroshi Okumura Nobuhiko Yamaguchi Improved Training of CAE-Based Defect Detectors Using Structural Noise Applied Sciences artificial intelligence image processing feature detection autoencoder convolutional autoencoder |
| title | Improved Training of CAE-Based Defect Detectors Using Structural Noise |
| title_full | Improved Training of CAE-Based Defect Detectors Using Structural Noise |
| title_fullStr | Improved Training of CAE-Based Defect Detectors Using Structural Noise |
| title_full_unstemmed | Improved Training of CAE-Based Defect Detectors Using Structural Noise |
| title_short | Improved Training of CAE-Based Defect Detectors Using Structural Noise |
| title_sort | improved training of cae based defect detectors using structural noise |
| topic | artificial intelligence image processing feature detection autoencoder convolutional autoencoder |
| url | https://www.mdpi.com/2076-3417/11/24/12062 |
| work_keys_str_mv | AT reinamurakami improvedtrainingofcaebaseddefectdetectorsusingstructuralnoise AT valentingrave improvedtrainingofcaebaseddefectdetectorsusingstructuralnoise AT osamufukuda improvedtrainingofcaebaseddefectdetectorsusingstructuralnoise AT hiroshiokumura improvedtrainingofcaebaseddefectdetectorsusingstructuralnoise AT nobuhikoyamaguchi improvedtrainingofcaebaseddefectdetectorsusingstructuralnoise |