SDSC-AI4Science seminar: "Unsupervised Domain Adaptation for Optoacoustic Imaging: Leveraging Unpaired Data for Improved Results"
![Thumbnail](/static/img/default.jpg)
Event details
Date | 25.04.2023 |
Hour | 16:00 › 17:00 |
Speaker | Dr. Firat Ozdemir, Sr. Data Scientist at SDSC |
Location | Online |
Category | Conferences - Seminars |
Event Language | English |
Abstract
With limited data or annotations, a common practice in AI is to build upon or exploit models pretrained on large-scale natural image datasets and/or generic natural language datasets. However, in domain sciences, such as optoacoustic imaging, it can be challenging to benefit from such pretrained models. Optoacoustic imaging involves the reconstruction of tissue images from acoustic signals generated by a light source. As a novel non-invasive tomographic imaging modality, optoacoustic imaging has a strong potential. In this talk, I will describe:
(i) optoacoustic imaging in simple terms,
(ii) some of the challenges that can be tackled with data science techniques,
(iii) supervised translation with limited-quality paired datasets,
(iv) improvements we achieved with unsupervised image-to-image translation using unpaired datasets (e.g., CycleGAN).
This is a joint work with the Multiscale Functional and Molecular Imaging Lab at ETH Zurich.
Organizers
The SDSC - AI4Science seminar is co-organized monthly by the EPFL AI4Science Initiative and the Swiss Data Science Center and focussing on projects in which data science, statistics, machine learning and AI are applied to the sciences. Each seminar will feature a presentation of one applied project, geared towards an audience with expertise in Data Science methods, from the initial formulation of a research question in science associated with sources of data, to the model, algorithms and analyses produced. The presentation will be highlighting the choices made, the challenges encountered, interesting technical questions and possible further developments. A number of the projects presented will be collaborative projects of the Swiss Data Science Center. One of the objectives of the seminar is to foster exchanges between researchers working in methods and applied data science research in the sciences, and to create new opportunities of collaborations. Each session will feature a talk followed by a discussion with the audience, to be continued over fingerfood and drinks.
With limited data or annotations, a common practice in AI is to build upon or exploit models pretrained on large-scale natural image datasets and/or generic natural language datasets. However, in domain sciences, such as optoacoustic imaging, it can be challenging to benefit from such pretrained models. Optoacoustic imaging involves the reconstruction of tissue images from acoustic signals generated by a light source. As a novel non-invasive tomographic imaging modality, optoacoustic imaging has a strong potential. In this talk, I will describe:
(i) optoacoustic imaging in simple terms,
(ii) some of the challenges that can be tackled with data science techniques,
(iii) supervised translation with limited-quality paired datasets,
(iv) improvements we achieved with unsupervised image-to-image translation using unpaired datasets (e.g., CycleGAN).
This is a joint work with the Multiscale Functional and Molecular Imaging Lab at ETH Zurich.
Organizers
The SDSC - AI4Science seminar is co-organized monthly by the EPFL AI4Science Initiative and the Swiss Data Science Center and focussing on projects in which data science, statistics, machine learning and AI are applied to the sciences. Each seminar will feature a presentation of one applied project, geared towards an audience with expertise in Data Science methods, from the initial formulation of a research question in science associated with sources of data, to the model, algorithms and analyses produced. The presentation will be highlighting the choices made, the challenges encountered, interesting technical questions and possible further developments. A number of the projects presented will be collaborative projects of the Swiss Data Science Center. One of the objectives of the seminar is to foster exchanges between researchers working in methods and applied data science research in the sciences, and to create new opportunities of collaborations. Each session will feature a talk followed by a discussion with the audience, to be continued over fingerfood and drinks.
Practical information
- Informed public
- Free