ALGM: Adaptive Local-then-Global Token Merging for Efficient Semantic Segmentation with Plain Vision Transformers
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929386646667264 |
|---|---|
| author | Norouzi, Narges Orlova, Svetlana de Geus, Daan Dubbelman, Gijs |
| author_facet | Norouzi, Narges Orlova, Svetlana de Geus, Daan Dubbelman, Gijs |
| contents | This work presents Adaptive Local-then-Global Merging (ALGM), a token reduction method for semantic segmentation networks that use plain Vision Transformers. ALGM merges tokens in two stages: (1) In the first network layer, it merges similar tokens within a small local window and (2) halfway through the network, it merges similar tokens across the entire image. This is motivated by an analysis in which we found that, in those situations, tokens with a high cosine similarity can likely be merged without a drop in segmentation quality. With extensive experiments across multiple datasets and network configurations, we show that ALGM not only significantly improves the throughput by up to 100%, but can also enhance the mean IoU by up to +1.1, thereby achieving a better trade-off between segmentation quality and efficiency than existing methods. Moreover, our approach is adaptive during inference, meaning that the same model can be used for optimal efficiency or accuracy, depending on the application. Code is available at https://tue-mps.github.io/ALGM. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_09936 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | ALGM: Adaptive Local-then-Global Token Merging for Efficient Semantic Segmentation with Plain Vision Transformers Norouzi, Narges Orlova, Svetlana de Geus, Daan Dubbelman, Gijs Computer Vision and Pattern Recognition This work presents Adaptive Local-then-Global Merging (ALGM), a token reduction method for semantic segmentation networks that use plain Vision Transformers. ALGM merges tokens in two stages: (1) In the first network layer, it merges similar tokens within a small local window and (2) halfway through the network, it merges similar tokens across the entire image. This is motivated by an analysis in which we found that, in those situations, tokens with a high cosine similarity can likely be merged without a drop in segmentation quality. With extensive experiments across multiple datasets and network configurations, we show that ALGM not only significantly improves the throughput by up to 100%, but can also enhance the mean IoU by up to +1.1, thereby achieving a better trade-off between segmentation quality and efficiency than existing methods. Moreover, our approach is adaptive during inference, meaning that the same model can be used for optimal efficiency or accuracy, depending on the application. Code is available at https://tue-mps.github.io/ALGM. |
| title | ALGM: Adaptive Local-then-Global Token Merging for Efficient Semantic Segmentation with Plain Vision Transformers |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2406.09936 |