Using Unreliable Pseudo-Labels for Label-Efficient Semantic Segmentation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2023
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909290915168256 |
|---|---|
| author | Wang, Haochen Wang, Yuchao Shen, Yujun Fan, Junsong Wang, Yuxi Zhang, Zhaoxiang |
| author_facet | Wang, Haochen Wang, Yuchao Shen, Yujun Fan, Junsong Wang, Yuxi Zhang, Zhaoxiang |
| contents | The crux of label-efficient semantic segmentation is to produce high-quality pseudo-labels to leverage a large amount of unlabeled or weakly labeled data. A common practice is to select the highly confident predictions as the pseudo-ground-truths for each pixel, but it leads to a problem that most pixels may be left unused due to their unreliability. However, we argue that every pixel matters to the model training, even those unreliable and ambiguous pixels. Intuitively, an unreliable prediction may get confused among the top classes, however, it should be confident about the pixel not belonging to the remaining classes. Hence, such a pixel can be convincingly treated as a negative key to those most unlikely categories. Therefore, we develop an effective pipeline to make sufficient use of unlabeled data. Concretely, we separate reliable and unreliable pixels via the entropy of predictions, push each unreliable pixel to a category-wise queue that consists of negative keys, and manage to train the model with all candidate pixels. Considering the training evolution, we adaptively adjust the threshold for the reliable-unreliable partition. Experimental results on various benchmarks and training settings demonstrate the superiority of our approach over the state-of-the-art alternatives. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2306_02314 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Using Unreliable Pseudo-Labels for Label-Efficient Semantic Segmentation Wang, Haochen Wang, Yuchao Shen, Yujun Fan, Junsong Wang, Yuxi Zhang, Zhaoxiang Computer Vision and Pattern Recognition The crux of label-efficient semantic segmentation is to produce high-quality pseudo-labels to leverage a large amount of unlabeled or weakly labeled data. A common practice is to select the highly confident predictions as the pseudo-ground-truths for each pixel, but it leads to a problem that most pixels may be left unused due to their unreliability. However, we argue that every pixel matters to the model training, even those unreliable and ambiguous pixels. Intuitively, an unreliable prediction may get confused among the top classes, however, it should be confident about the pixel not belonging to the remaining classes. Hence, such a pixel can be convincingly treated as a negative key to those most unlikely categories. Therefore, we develop an effective pipeline to make sufficient use of unlabeled data. Concretely, we separate reliable and unreliable pixels via the entropy of predictions, push each unreliable pixel to a category-wise queue that consists of negative keys, and manage to train the model with all candidate pixels. Considering the training evolution, we adaptively adjust the threshold for the reliable-unreliable partition. Experimental results on various benchmarks and training settings demonstrate the superiority of our approach over the state-of-the-art alternatives. |
| title | Using Unreliable Pseudo-Labels for Label-Efficient Semantic Segmentation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2306.02314 |