Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection
Image-based rail defect detection could be conceptually defined as an object detection task in computer vision. However, unlike academic object detection tasks, this practical industrial application suffers from two unique challenges, including object ambiguity and insufficient annotations. To overc...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-08-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/12/16/8006 |
| _version_ | 1797439763457769472 |
|---|---|
| author | Hongjue Li Hailang Li Zhixiong Hou Haoran Song Junbo Liu Peng Dai |
| author_facet | Hongjue Li Hailang Li Zhixiong Hou Haoran Song Junbo Liu Peng Dai |
| author_sort | Hongjue Li |
| collection | DOAJ |
| description | Image-based rail defect detection could be conceptually defined as an object detection task in computer vision. However, unlike academic object detection tasks, this practical industrial application suffers from two unique challenges, including object ambiguity and insufficient annotations. To overcome these challenges, we introduce the pixel-wise attention mechanism to fully exploit features of annotated defects, and develop a feature augmentation framework to tackle the defect detection problem. The pixel-wise attention is conducted through a learnable pixel-level similarity between input and support features to obtain augmented features. These augmented features contain co-existing information from input images and multi-class support defects. The final output features are augmented and refined by support features, thus endowing the model to distinguish between ambiguous defect patterns based on insufficient annotated samples. Experiments on the rail defect dataset demonstrate that feature augmentation can help balance the sensitivity and robustness of the model. On our collected dataset with eight defected classes, our algorithm achieves 11.32% higher mAP@.5 compared with original YOLOv5 and 4.27% higher mAP@.5 compared with Faster R-CNN. |
| first_indexed | 2024-03-09T11:57:49Z |
| format | Article |
| id | doaj.art-992dde5cf379446f86e7de4042143d56 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T11:57:49Z |
| publishDate | 2022-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-992dde5cf379446f86e7de4042143d562023-11-30T23:07:16ZengMDPI AGApplied Sciences2076-34172022-08-011216800610.3390/app12168006Feature Augmentation Based on Pixel-Wise Attention for Rail Defect DetectionHongjue Li0Hailang Li1Zhixiong Hou2Haoran Song3Junbo Liu4Peng Dai5School of Astronautics, Beihang University, Beijing 100191, ChinaInfrastructure Inspection Research Institute, China Academy of Railway Science Corporation Limited, Beijing 100081, ChinaInfrastructure Inspection Research Institute, China Academy of Railway Science Corporation Limited, Beijing 100081, ChinaInfrastructure Inspection Research Institute, China Academy of Railway Science Corporation Limited, Beijing 100081, ChinaInfrastructure Inspection Research Institute, China Academy of Railway Science Corporation Limited, Beijing 100081, ChinaInfrastructure Inspection Research Institute, China Academy of Railway Science Corporation Limited, Beijing 100081, ChinaImage-based rail defect detection could be conceptually defined as an object detection task in computer vision. However, unlike academic object detection tasks, this practical industrial application suffers from two unique challenges, including object ambiguity and insufficient annotations. To overcome these challenges, we introduce the pixel-wise attention mechanism to fully exploit features of annotated defects, and develop a feature augmentation framework to tackle the defect detection problem. The pixel-wise attention is conducted through a learnable pixel-level similarity between input and support features to obtain augmented features. These augmented features contain co-existing information from input images and multi-class support defects. The final output features are augmented and refined by support features, thus endowing the model to distinguish between ambiguous defect patterns based on insufficient annotated samples. Experiments on the rail defect dataset demonstrate that feature augmentation can help balance the sensitivity and robustness of the model. On our collected dataset with eight defected classes, our algorithm achieves 11.32% higher mAP@.5 compared with original YOLOv5 and 4.27% higher mAP@.5 compared with Faster R-CNN.https://www.mdpi.com/2076-3417/12/16/8006object detectionpixel-wise attentionfeature augmentationrail defect |
| spellingShingle | Hongjue Li Hailang Li Zhixiong Hou Haoran Song Junbo Liu Peng Dai Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection Applied Sciences object detection pixel-wise attention feature augmentation rail defect |
| title | Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection |
| title_full | Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection |
| title_fullStr | Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection |
| title_full_unstemmed | Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection |
| title_short | Feature Augmentation Based on Pixel-Wise Attention for Rail Defect Detection |
| title_sort | feature augmentation based on pixel wise attention for rail defect detection |
| topic | object detection pixel-wise attention feature augmentation rail defect |
| url | https://www.mdpi.com/2076-3417/12/16/8006 |
| work_keys_str_mv | AT hongjueli featureaugmentationbasedonpixelwiseattentionforraildefectdetection AT hailangli featureaugmentationbasedonpixelwiseattentionforraildefectdetection AT zhixionghou featureaugmentationbasedonpixelwiseattentionforraildefectdetection AT haoransong featureaugmentationbasedonpixelwiseattentionforraildefectdetection AT junboliu featureaugmentationbasedonpixelwiseattentionforraildefectdetection AT pengdai featureaugmentationbasedonpixelwiseattentionforraildefectdetection |