A Min-Max Based Hyperparameter Estimation For Domain-Adapted Segmentation Of Amoeboid Cells
| Autor: | Elisabeth Labruyère, Rituparna Sarkar, Jean-Christophe Olivo-Marin, Suvadip Mukherjee |
|---|---|
| Přispěvatelé: | Analyse d'images biologiques - Biological Image Analysis (BIA), Institut Pasteur [Paris] (IP)-Centre National de la Recherche Scientifique (CNRS), ANR-16-CONV-0005,INCEPTION,Institut Convergences pour l'étude de l'Emergence des Pathologies au Travers des Individus et des populatiONs(2016) |
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Hyperparameter
0303 health sciences Artificial neural network business.industry Computer science Pattern recognition Image segmentation [SDV.BIBS]Life Sciences [q-bio]/Quantitative Methods [q-bio.QM] Synthetic data Expression (mathematics) 030218 nuclear medicine & medical imaging Domain (software engineering) 03 medical and health sciences 0302 clinical medicine [SDV.BC.IC]Life Sciences [q-bio]/Cellular Biology/Cell Behavior [q-bio.CB] Segmentation Penalty method [SDV.MP.PAR]Life Sciences [q-bio]/Microbiology and Parasitology/Parasitology Artificial intelligence business 030304 developmental biology |
| Zdroj: | 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI) 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), Apr 2021, Nice, France. pp.1869-1872, ⟨10.1109/ISBI48211.2021.9433864⟩ ISBI |
| DOI: | 10.1109/ISBI48211.2021.9433864⟩ |
| Popis: | Domain adaption is a tool to fit models trained on a source dataset to characteristically different target samples. During training, such methods typically seek to minimize the segmentation loss, in conjunction with a penalty function for domain misalignment. State-of-the-art solutions are sensitive to heuristically chosen hyper-parameters that dictate the proportion of the two cost functions. We address this issue by introducing a novel strategy for hyper-parameter estimation via min-max optimization of the deep neural network’s loss function. Our solution provides an analytical expression for the model hyper-parameters, which are iteratively estimated during training. Experimental evaluation on both synthetic data and microscopy images of ameoboid cells Entamoeba histolytica attest to the effectiveness of our solution for deep domain-adapted segmentation in bioimaging. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|