Joint training of generic CNN-CRF models with stochastic optimization
We propose a new CNN-CRF end-to-end learning framework, which is based on joint stochastic optimization with respect to both Convolutional Neural Network (CNN) and Conditional Random Field (CRF) parameters. While stochastic gradient descent is a standard technique for CNN training, it was not used f...
| Main Authors: | , , , , , |
|---|---|
| פורמט: | Conference item |
| יצא לאור: |
Springer, Cham
2017
|
| _version_ | 1826303962247069696 |
|---|---|
| author | Kirillov, A Schlesinger, D Zheng, S Savchynskyy, B Torr, P Rother, C |
| author_facet | Kirillov, A Schlesinger, D Zheng, S Savchynskyy, B Torr, P Rother, C |
| author_sort | Kirillov, A |
| collection | OXFORD |
| description | We propose a new CNN-CRF end-to-end learning framework, which is based on joint stochastic optimization with respect to both Convolutional Neural Network (CNN) and Conditional Random Field (CRF) parameters. While stochastic gradient descent is a standard technique for CNN training, it was not used for joint models so far. We show that our learning method is (i) general, i.e. it applies to arbitrary CNN and CRF architectures and potential functions; (ii) scalable, i.e. it has a low memory footprint and straightforwardly parallelizes on GPUs; (iii) easy in implementation. Additionally, the unified CNN-CRF optimization approach simplifies a potential hardware implementation. We empirically evaluate our method on the task of semantic labeling of body parts in depth images and show that it compares favorably to competing techniques. |
| first_indexed | 2024-03-07T06:10:38Z |
| format | Conference item |
| id | oxford-uuid:ef5c8eae-6e03-41f4-bd0c-5e5b934f97eb |
| institution | University of Oxford |
| last_indexed | 2024-03-07T06:10:38Z |
| publishDate | 2017 |
| publisher | Springer, Cham |
| record_format | dspace |
| spelling | oxford-uuid:ef5c8eae-6e03-41f4-bd0c-5e5b934f97eb2022-03-27T11:39:39ZJoint training of generic CNN-CRF models with stochastic optimizationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:ef5c8eae-6e03-41f4-bd0c-5e5b934f97ebSymplectic Elements at OxfordSpringer, Cham2017Kirillov, ASchlesinger, DZheng, SSavchynskyy, BTorr, PRother, CWe propose a new CNN-CRF end-to-end learning framework, which is based on joint stochastic optimization with respect to both Convolutional Neural Network (CNN) and Conditional Random Field (CRF) parameters. While stochastic gradient descent is a standard technique for CNN training, it was not used for joint models so far. We show that our learning method is (i) general, i.e. it applies to arbitrary CNN and CRF architectures and potential functions; (ii) scalable, i.e. it has a low memory footprint and straightforwardly parallelizes on GPUs; (iii) easy in implementation. Additionally, the unified CNN-CRF optimization approach simplifies a potential hardware implementation. We empirically evaluate our method on the task of semantic labeling of body parts in depth images and show that it compares favorably to competing techniques. |
| spellingShingle | Kirillov, A Schlesinger, D Zheng, S Savchynskyy, B Torr, P Rother, C Joint training of generic CNN-CRF models with stochastic optimization |
| title | Joint training of generic CNN-CRF models with stochastic optimization |
| title_full | Joint training of generic CNN-CRF models with stochastic optimization |
| title_fullStr | Joint training of generic CNN-CRF models with stochastic optimization |
| title_full_unstemmed | Joint training of generic CNN-CRF models with stochastic optimization |
| title_short | Joint training of generic CNN-CRF models with stochastic optimization |
| title_sort | joint training of generic cnn crf models with stochastic optimization |
| work_keys_str_mv | AT kirillova jointtrainingofgenericcnncrfmodelswithstochasticoptimization AT schlesingerd jointtrainingofgenericcnncrfmodelswithstochasticoptimization AT zhengs jointtrainingofgenericcnncrfmodelswithstochasticoptimization AT savchynskyyb jointtrainingofgenericcnncrfmodelswithstochasticoptimization AT torrp jointtrainingofgenericcnncrfmodelswithstochasticoptimization AT rotherc jointtrainingofgenericcnncrfmodelswithstochasticoptimization |