Speeding up convolutional neural networks with low rank expansions
The focus of this paper is speeding up the application of convolutional neural networks. While delivering impressive results across a range of computer vision and machine learning tasks, these networks are computationally demanding, limiting their deployability. Convolutional layers generally consum...
| Main Authors: | , , |
|---|---|
| 格式: | Conference item |
| 出版: |
BMVA Press
2014
|
| _version_ | 1826315169850982400 |
|---|---|
| author | Jaderberg, M Vedaldi, A Zisserman, A |
| author_facet | Jaderberg, M Vedaldi, A Zisserman, A |
| author_sort | Jaderberg, M |
| collection | OXFORD |
| description | The focus of this paper is speeding up the application of convolutional neural networks. While delivering impressive results across a range of computer vision and machine learning tasks, these networks are computationally demanding, limiting their deployability. Convolutional layers generally consume the bulk of the processing time, and so in this work we present two simple schemes for drastically speeding up these layers. This is achieved by exploiting cross-channel or filter redundancy to construct a low rank basis of filters that are rank-1 in the spatial domain. Our methods are architecture agnostic, and can be easily applied to existing CPU and GPU convolutional frameworks for tuneable speedup performance. We demonstrate this with a real world network designed for scene text character recognition [15], showing a possible 2.5× speedup with no loss in accuracy, and 4.5× speedup with less than 1% drop in accuracy, still achieving state-of-the-art on standard benchmarks. |
| first_indexed | 2024-03-07T04:29:48Z |
| format | Conference item |
| id | oxford-uuid:cdeb70cb-0b3f-4e99-9e9c-2ce50f8bb8f8 |
| institution | University of Oxford |
| last_indexed | 2024-12-09T03:20:46Z |
| publishDate | 2014 |
| publisher | BMVA Press |
| record_format | dspace |
| spelling | oxford-uuid:cdeb70cb-0b3f-4e99-9e9c-2ce50f8bb8f82024-11-05T13:52:28ZSpeeding up convolutional neural networks with low rank expansionsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:cdeb70cb-0b3f-4e99-9e9c-2ce50f8bb8f8Symplectic Elements at OxfordBMVA Press2014Jaderberg, MVedaldi, AZisserman, AThe focus of this paper is speeding up the application of convolutional neural networks. While delivering impressive results across a range of computer vision and machine learning tasks, these networks are computationally demanding, limiting their deployability. Convolutional layers generally consume the bulk of the processing time, and so in this work we present two simple schemes for drastically speeding up these layers. This is achieved by exploiting cross-channel or filter redundancy to construct a low rank basis of filters that are rank-1 in the spatial domain. Our methods are architecture agnostic, and can be easily applied to existing CPU and GPU convolutional frameworks for tuneable speedup performance. We demonstrate this with a real world network designed for scene text character recognition [15], showing a possible 2.5× speedup with no loss in accuracy, and 4.5× speedup with less than 1% drop in accuracy, still achieving state-of-the-art on standard benchmarks. |
| spellingShingle | Jaderberg, M Vedaldi, A Zisserman, A Speeding up convolutional neural networks with low rank expansions |
| title | Speeding up convolutional neural networks with low rank expansions |
| title_full | Speeding up convolutional neural networks with low rank expansions |
| title_fullStr | Speeding up convolutional neural networks with low rank expansions |
| title_full_unstemmed | Speeding up convolutional neural networks with low rank expansions |
| title_short | Speeding up convolutional neural networks with low rank expansions |
| title_sort | speeding up convolutional neural networks with low rank expansions |
| work_keys_str_mv | AT jaderbergm speedingupconvolutionalneuralnetworkswithlowrankexpansions AT vedaldia speedingupconvolutionalneuralnetworkswithlowrankexpansions AT zissermana speedingupconvolutionalneuralnetworkswithlowrankexpansions |