The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts
© 2019 SPIE. In a recent paper [Goy et al., Phys. Rev. Lett. 121, 243902, 2018], we showed that deep neural networks (DNNs) are very efficient solvers for phase retrieval problems, especially when the photon budget is limited. However, the performance of the DNN is strongly conditioned by a preproce...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SPIE
2021
|
| Online Access: | https://hdl.handle.net/1721.1/136711 |
| _version_ | 1826191844129636352 |
|---|---|
| author | Goy, Alexandre Arthur, Kwabena Li, Shuai Barbastathis, George |
| author_facet | Goy, Alexandre Arthur, Kwabena Li, Shuai Barbastathis, George |
| author_sort | Goy, Alexandre |
| collection | MIT |
| description | © 2019 SPIE. In a recent paper [Goy et al., Phys. Rev. Lett. 121, 243902, 2018], we showed that deep neural networks (DNNs) are very efficient solvers for phase retrieval problems, especially when the photon budget is limited. However, the performance of the DNN is strongly conditioned by a preprocessing step that consists in producing a proper initial guess. In this paper, we study the influence of the preprocessing in more details, in particular the choice of the preprocessing operator. We also empirically demonstrate that, for a DenseNet architecture, the performance of the DNN increases with the number of layers up to a point after which it saturates. |
| first_indexed | 2024-09-23T09:02:09Z |
| format | Article |
| id | mit-1721.1/136711 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T09:02:09Z |
| publishDate | 2021 |
| publisher | SPIE |
| record_format | dspace |
| spelling | mit-1721.1/1367112021-10-29T03:38:15Z The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts Goy, Alexandre Arthur, Kwabena Li, Shuai Barbastathis, George © 2019 SPIE. In a recent paper [Goy et al., Phys. Rev. Lett. 121, 243902, 2018], we showed that deep neural networks (DNNs) are very efficient solvers for phase retrieval problems, especially when the photon budget is limited. However, the performance of the DNN is strongly conditioned by a preprocessing step that consists in producing a proper initial guess. In this paper, we study the influence of the preprocessing in more details, in particular the choice of the preprocessing operator. We also empirically demonstrate that, for a DenseNet architecture, the performance of the DNN increases with the number of layers up to a point after which it saturates. 2021-10-28T15:26:52Z 2021-10-28T15:26:52Z 2019-03-04 2020-06-22T18:57:28Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/136711 Goy, Alexandre, Arthur, Kwabena, Li, Shuai and Barbastathis, George. 2019. "The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts." Progress in Biomedical Optics and Imaging - Proceedings of SPIE, 10887. en 10.1117/12.2513314 Progress in Biomedical Optics and Imaging - Proceedings of SPIE Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf SPIE SPIE |
| spellingShingle | Goy, Alexandre Arthur, Kwabena Li, Shuai Barbastathis, George The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title | The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title_full | The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title_fullStr | The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title_full_unstemmed | The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title_short | The importance of physical pre-processors for quantitative phase retrieval under extremely low photon counts |
| title_sort | importance of physical pre processors for quantitative phase retrieval under extremely low photon counts |
| url | https://hdl.handle.net/1721.1/136711 |
| work_keys_str_mv | AT goyalexandre theimportanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT arthurkwabena theimportanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT lishuai theimportanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT barbastathisgeorge theimportanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT goyalexandre importanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT arthurkwabena importanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT lishuai importanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts AT barbastathisgeorge importanceofphysicalpreprocessorsforquantitativephaseretrievalunderextremelylowphotoncounts |