Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times.
<p>In each replicate, a few (1, 3, 5, 10, 20, 30, or 50) images are randomly selected from the benchmark dataset, and <i>λ</i> is tuned for each image to a proper value so that MCount gives the same counting number as the label. Then, the average of <i>λ</i> is chosen f...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | , , , |
| Published: |
2025
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1852021995939561472 |
|---|---|
| author | Sijie Chen (1622476) |
| author2 | Po-Hsun Huang (183737) Hyungseok Kim (14635448) Yuhe Cui (20902959) Cullen R. Buie (471587) |
| author2_role | author author author author |
| author_facet | Sijie Chen (1622476) Po-Hsun Huang (183737) Hyungseok Kim (14635448) Yuhe Cui (20902959) Cullen R. Buie (471587) |
| author_role | author |
| dc.creator.none.fl_str_mv | Sijie Chen (1622476) Po-Hsun Huang (183737) Hyungseok Kim (14635448) Yuhe Cui (20902959) Cullen R. Buie (471587) |
| dc.date.none.fl_str_mv | 2025-03-19T17:33:44Z |
| dc.identifier.none.fl_str_mv | 10.1371/journal.pone.0311242.g005 |
| dc.relation.none.fl_str_mv | https://figshare.com/articles/figure/Distribution_of_average_recognition_error_rates_on_the_benchmark_dataset_when_repeatedly_implementing_the_hyperparameter-tuning_method_using_i_n_i_labeled_images_1000_times_/28626564 |
| dc.rights.none.fl_str_mv | CC BY 4.0 info:eu-repo/semantics/openAccess |
| dc.subject.none.fl_str_mv | Biophysics Biochemistry Molecular Biology Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Chemical Sciences not elsewhere classified Information Systems not elsewhere classified escherichia coli </ average error rate assessing microbial growth requires two hyperparameters regional candidate circles propose statistical methods efficient colony counting limited labeled data div >< p 54 %), autocellseg accurate colony counting colony counting 54 %), regional algorithms hyperparameters using propose mcount throughput workflows throughput plating promising solution precisely labeled merged colonies known solution facilitate deployment contour information common occurrence accurately infer 960 images 847 segments 31 %). |
| dc.title.none.fl_str_mv | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| dc.type.none.fl_str_mv | Image Figure info:eu-repo/semantics/publishedVersion image |
| description | <p>In each replicate, a few (1, 3, 5, 10, 20, 30, or 50) images are randomly selected from the benchmark dataset, and <i>λ</i> is tuned for each image to a proper value so that MCount gives the same counting number as the label. Then, the average of <i>λ</i> is chosen for this replicate, and the average recognition error rate is calculated on the benchmark using this <i>λ</i>. By simulating this procedure for 1000 replicates, we can plot the distribution of average recognition error rates. As expected, increasing <i>n</i> results in a narrower distribution of average recognition error rates, leading to more consistent performance. Note that when , all recognition errors fall in the range of 3.5% to 13% with a mean of 5.17% (median of 4.77%), much lower than the recognition error rate of NICE at 16.54% (15.79%).</p> |
| eu_rights_str_mv | openAccess |
| id | Manara_6777ff630c8a3ddef96cfceedea39a7c |
| identifier_str_mv | 10.1371/journal.pone.0311242.g005 |
| network_acronym_str | Manara |
| network_name_str | ManaraRepo |
| oai_identifier_str | oai:figshare.com:article/28626564 |
| publishDate | 2025 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| rights_invalid_str_mv | CC BY 4.0 |
| spelling | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times.Sijie Chen (1622476)Po-Hsun Huang (183737)Hyungseok Kim (14635448)Yuhe Cui (20902959)Cullen R. Buie (471587)BiophysicsBiochemistryMolecular BiologyBiological Sciences not elsewhere classifiedMathematical Sciences not elsewhere classifiedChemical Sciences not elsewhere classifiedInformation Systems not elsewhere classifiedescherichia coli </average error rateassessing microbial growthrequires two hyperparametersregional candidate circlespropose statistical methodsefficient colony countinglimited labeled datadiv >< p54 %), autocellsegaccurate colony countingcolony counting54 %),regional algorithmshyperparameters usingpropose mcountthroughput workflowsthroughput platingpromising solutionprecisely labeledmerged coloniesknown solutionfacilitate deploymentcontour informationcommon occurrenceaccurately infer960 images847 segments31 %).<p>In each replicate, a few (1, 3, 5, 10, 20, 30, or 50) images are randomly selected from the benchmark dataset, and <i>λ</i> is tuned for each image to a proper value so that MCount gives the same counting number as the label. Then, the average of <i>λ</i> is chosen for this replicate, and the average recognition error rate is calculated on the benchmark using this <i>λ</i>. By simulating this procedure for 1000 replicates, we can plot the distribution of average recognition error rates. As expected, increasing <i>n</i> results in a narrower distribution of average recognition error rates, leading to more consistent performance. Note that when , all recognition errors fall in the range of 3.5% to 13% with a mean of 5.17% (median of 4.77%), much lower than the recognition error rate of NICE at 16.54% (15.79%).</p>2025-03-19T17:33:44ZImageFigureinfo:eu-repo/semantics/publishedVersionimage10.1371/journal.pone.0311242.g005https://figshare.com/articles/figure/Distribution_of_average_recognition_error_rates_on_the_benchmark_dataset_when_repeatedly_implementing_the_hyperparameter-tuning_method_using_i_n_i_labeled_images_1000_times_/28626564CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/286265642025-03-19T17:33:44Z |
| spellingShingle | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. Sijie Chen (1622476) Biophysics Biochemistry Molecular Biology Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Chemical Sciences not elsewhere classified Information Systems not elsewhere classified escherichia coli </ average error rate assessing microbial growth requires two hyperparameters regional candidate circles propose statistical methods efficient colony counting limited labeled data div >< p 54 %), autocellseg accurate colony counting colony counting 54 %), regional algorithms hyperparameters using propose mcount throughput workflows throughput plating promising solution precisely labeled merged colonies known solution facilitate deployment contour information common occurrence accurately infer 960 images 847 segments 31 %). |
| status_str | publishedVersion |
| title | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| title_full | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| title_fullStr | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| title_full_unstemmed | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| title_short | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| title_sort | Distribution of average recognition error rates on the benchmark dataset when repeatedly implementing the hyperparameter-tuning method using <i>n</i> labeled images 1000 times. |
| topic | Biophysics Biochemistry Molecular Biology Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Chemical Sciences not elsewhere classified Information Systems not elsewhere classified escherichia coli </ average error rate assessing microbial growth requires two hyperparameters regional candidate circles propose statistical methods efficient colony counting limited labeled data div >< p 54 %), autocellseg accurate colony counting colony counting 54 %), regional algorithms hyperparameters using propose mcount throughput workflows throughput plating promising solution precisely labeled merged colonies known solution facilitate deployment contour information common occurrence accurately infer 960 images 847 segments 31 %). |