Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices: empirical study from transfer learning to optimization
Despite the emergence of mobile health and the success of deep learning (DL), deploying production-ready DL models to resource-limited devices remains challenging. Especially, during inference time, the speed of DL models becomes relevant. We aimed to accelerate inference time for Gram-stained analy...
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Dokumenttyp: | Article (Journal) |
| Sprache: | Englisch |
| Veröffentlicht: |
4 November 2022
|
| In: |
Biomedicines
Year: 2022, Jahrgang: 10, Heft: 11, Pages: 1-12 |
| ISSN: | 2227-9059 |
| DOI: | 10.3390/biomedicines10112808 |
| Online-Zugang: | Verlag, kostenfrei, Volltext: https://doi.org/10.3390/biomedicines10112808 Verlag, kostenfrei, Volltext: https://www.mdpi.com/2227-9059/10/11/2808 
|
| Verfasserangaben: | Hee E. Kim, Mate E. Maros, Fabian Siegel, Thomas Ganslandt |
MARC
| LEADER | 00000caa a2200000 c 4500 | ||
|---|---|---|---|
| 001 | 1852793805 | ||
| 003 | DE-627 | ||
| 005 | 20240306113537.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 230717s2022 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.3390/biomedicines10112808 |2 doi | |
| 035 | |a (DE-627)1852793805 | ||
| 035 | |a (DE-599)KXP1852793805 | ||
| 035 | |a (OCoLC)1425062419 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rda | ||
| 041 | |a eng | ||
| 084 | |a 33 |2 sdnb | ||
| 100 | 1 | |a Kim, Hee Eun |d 1986- |e VerfasserIn |0 (DE-588)1296369323 |0 (DE-627)1852791470 |4 aut | |
| 245 | 1 | 0 | |a Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices |b empirical study from transfer learning to optimization |c Hee E. Kim, Mate E. Maros, Fabian Siegel, Thomas Ganslandt |
| 264 | 1 | |c 4 November 2022 | |
| 300 | |b Illustrationen | ||
| 300 | |a 12 | ||
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a Computermedien |b c |2 rdamedia | ||
| 338 | |a Online-Ressource |b cr |2 rdacarrier | ||
| 500 | |a Gesehen am 17.07.2023 | ||
| 520 | |a Despite the emergence of mobile health and the success of deep learning (DL), deploying production-ready DL models to resource-limited devices remains challenging. Especially, during inference time, the speed of DL models becomes relevant. We aimed to accelerate inference time for Gram-stained analysis, which is a tedious and manual task involving microorganism detection on whole slide images. Three DL models were optimized in three steps: transfer learning, pruning and quantization and then evaluated on two Android smartphones. Most convolutional layers (≥80%) had to be retrained for adaptation to the Gram-stained classification task. The combination of pruning and quantization demonstrated its utility to reduce the model size and inference time without compromising model quality. Pruning mainly contributed to model size reduction by 15×, while quantization reduced inference time by 3× and decreased model size by 4×. The combination of two reduced the baseline model by an overall factor of 46×. Optimized models were smaller than 6 MB and were able to process one image in <0.6 s on a Galaxy S10. Our findings demonstrate that methods for model compression are highly relevant for the successful deployment of DL solutions to resource-limited devices. | ||
| 650 | 4 | |a deep learning | |
| 650 | 4 | |a Gram-stained classification | |
| 650 | 4 | |a mHealth | |
| 650 | 4 | |a model compression | |
| 650 | 4 | |a pruning | |
| 650 | 4 | |a quantization | |
| 650 | 4 | |a rapid inference time | |
| 700 | 1 | |a Maros, Máté E. |d 1986- |e VerfasserIn |0 (DE-588)1144379407 |0 (DE-627)1004715153 |0 (DE-576)495364827 |4 aut | |
| 700 | 1 | |a Siegel, Fabian |d 1979- |e VerfasserIn |0 (DE-588)1034567551 |0 (DE-627)745868037 |0 (DE-576)382210514 |4 aut | |
| 700 | 1 | |a Ganslandt, Thomas |d 1969- |e VerfasserIn |0 (DE-588)124367720 |0 (DE-627)08581623X |0 (DE-576)294139915 |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t Biomedicines |d Basel : MDPI, 2013 |g 10(2022), 11, Artikel-ID 2808, Seite 1-12 |h Online-Ressource |w (DE-627)750370483 |w (DE-600)2720867-9 |w (DE-576)384589596 |x 2227-9059 |7 nnas |a Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices empirical study from transfer learning to optimization |
| 773 | 1 | 8 | |g volume:10 |g year:2022 |g number:11 |g elocationid:2808 |g pages:1-12 |g extent:12 |a Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices empirical study from transfer learning to optimization |
| 856 | 4 | 0 | |u https://doi.org/10.3390/biomedicines10112808 |x Verlag |x Resolving-System |z kostenfrei |3 Volltext |
| 856 | 4 | 0 | |u https://www.mdpi.com/2227-9059/10/11/2808 |x Verlag |z kostenfrei |3 Volltext |
| 951 | |a AR | ||
| 992 | |a 20230717 | ||
| 993 | |a Article | ||
| 994 | |a 2022 | ||
| 998 | |g 124367720 |a Ganslandt, Thomas |m 124367720:Ganslandt, Thomas |d 60000 |e 60000PG124367720 |k 0/60000/ |p 4 |y j | ||
| 998 | |g 1034567551 |a Siegel, Fabian |m 1034567551:Siegel, Fabian |d 60000 |d 63100 |e 60000PS1034567551 |e 63100PS1034567551 |k 0/60000/ |k 1/60000/63100/ |p 3 | ||
| 998 | |g 1144379407 |a Maros, Máté E. |m 1144379407:Maros, Máté E. |d 60000 |d 63000 |e 60000PM1144379407 |e 63000PM1144379407 |k 0/60000/ |k 1/60000/63000/ |p 2 | ||
| 998 | |g 1296369323 |a Kim, Hee Eun |m 1296369323:Kim, Hee Eun |d 60000 |d 65300 |e 60000PK1296369323 |e 65300PK1296369323 |k 0/60000/ |k 1/60000/65300/ |p 1 |x j | ||
| 999 | |a KXP-PPN1852793805 |e 4354229374 | ||
| BIB | |a Y | ||
| SER | |a journal | ||
| JSO | |a {"relHost":[{"title":[{"subtitle":"open access journal","title":"Biomedicines","title_sort":"Biomedicines"}],"note":["Gesehen am 12.08.20"],"type":{"bibl":"periodical","media":"Online-Ressource"},"pubHistory":["1.2013 -"],"language":["eng"],"physDesc":[{"extent":"Online-Ressource"}],"origin":[{"publisher":"MDPI","dateIssuedKey":"2013","dateIssuedDisp":"2013-","publisherPlace":"Basel"}],"id":{"issn":["2227-9059"],"zdb":["2720867-9"],"eki":["750370483"]},"part":{"year":"2022","issue":"11","extent":"12","text":"10(2022), 11, Artikel-ID 2808, Seite 1-12","volume":"10","pages":"1-12"},"recId":"750370483","disp":"Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices empirical study from transfer learning to optimizationBiomedicines"}],"note":["Gesehen am 17.07.2023"],"recId":"1852793805","name":{"displayForm":["Hee E. Kim, Mate E. Maros, Fabian Siegel, Thomas Ganslandt"]},"origin":[{"dateIssuedDisp":"4 November 2022","dateIssuedKey":"2022"}],"id":{"eki":["1852793805"],"doi":["10.3390/biomedicines10112808"]},"physDesc":[{"noteIll":"Illustrationen","extent":"12 S."}],"language":["eng"],"type":{"media":"Online-Ressource","bibl":"article-journal"},"title":[{"subtitle":"empirical study from transfer learning to optimization","title":"Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices","title_sort":"Rapid convolutional neural networks for Gram-stained image classification at inference time on mobile devices"}],"person":[{"display":"Kim, Hee Eun","role":"aut","given":"Hee Eun","family":"Kim"},{"display":"Maros, Máté E.","family":"Maros","given":"Máté E.","role":"aut"},{"display":"Siegel, Fabian","given":"Fabian","role":"aut","family":"Siegel"},{"display":"Ganslandt, Thomas","family":"Ganslandt","role":"aut","given":"Thomas"}]} | ||
| SRT | |a KIMHEEEUNMRAPIDCONVO4202 | ||