Semi-Automated Training of AI Vision Models
The adoption of AI vision models in specialized industries is often hindered by the substantial requirement for extensive, manually annotated image datasets. Even when employing transfer learning, robust model development typically necessitates tens of thousands of such images, a process that is tim...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-07-01
|
| Series: | AgriEngineering |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2624-7402/7/7/225 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850067586267480064 |
|---|---|
| author | Mathew G. Pelletier John D. Wanjura Greg A. Holt |
| author_facet | Mathew G. Pelletier John D. Wanjura Greg A. Holt |
| author_sort | Mathew G. Pelletier |
| collection | DOAJ |
| description | The adoption of AI vision models in specialized industries is often hindered by the substantial requirement for extensive, manually annotated image datasets. Even when employing transfer learning, robust model development typically necessitates tens of thousands of such images, a process that is time-consuming, costly, and demands consistent expert annotation. This technical note introduces a semi-automated method to significantly reduce this annotation burden. The proposed approach utilizes two general-purpose vision-transformer-to-caption (GP-ViTC) models to generate descriptive text from images. These captions are then processed by a custom-developed semantic classifier (SC), which requires only minimal training to predict the correct image class. This GP-ViTC + SC system demonstrated exemplary classification rates in test cases and can subsequently be used to automatically annotate large image datasets. While the inference speed of the GP-ViTC models is not suited for real-time applications (approximately 10 s per image), this method substantially lessens the labor and expertise required for dataset creation, thereby facilitating the development of new, high-speed, custom AI vision models for niche applications. This work details the approach and its successful application, offering a cost-effective pathway for generating tailored image training sets. |
| format | Article |
| id | doaj-art-bfa4e81413804730b2bf03cdb0ee5297 |
| institution | DOAJ |
| issn | 2624-7402 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | AgriEngineering |
| spelling | doaj-art-bfa4e81413804730b2bf03cdb0ee52972025-08-20T02:48:16ZengMDPI AGAgriEngineering2624-74022025-07-017722510.3390/agriengineering7070225Semi-Automated Training of AI Vision ModelsMathew G. Pelletier0John D. Wanjura1Greg A. Holt2Lubbock Gin-Laboratory, Cotton Production and Processing Research Unit, United States Department of Agriculture, Agricultural Research Services, Lubbock, TX 79403, USALubbock Gin-Laboratory, Cotton Production and Processing Research Unit, United States Department of Agriculture, Agricultural Research Services, Lubbock, TX 79403, USALubbock Gin-Laboratory, Cotton Production and Processing Research Unit, United States Department of Agriculture, Agricultural Research Services, Lubbock, TX 79403, USAThe adoption of AI vision models in specialized industries is often hindered by the substantial requirement for extensive, manually annotated image datasets. Even when employing transfer learning, robust model development typically necessitates tens of thousands of such images, a process that is time-consuming, costly, and demands consistent expert annotation. This technical note introduces a semi-automated method to significantly reduce this annotation burden. The proposed approach utilizes two general-purpose vision-transformer-to-caption (GP-ViTC) models to generate descriptive text from images. These captions are then processed by a custom-developed semantic classifier (SC), which requires only minimal training to predict the correct image class. This GP-ViTC + SC system demonstrated exemplary classification rates in test cases and can subsequently be used to automatically annotate large image datasets. While the inference speed of the GP-ViTC models is not suited for real-time applications (approximately 10 s per image), this method substantially lessens the labor and expertise required for dataset creation, thereby facilitating the development of new, high-speed, custom AI vision models for niche applications. This work details the approach and its successful application, offering a cost-effective pathway for generating tailored image training sets.https://www.mdpi.com/2624-7402/7/7/225machine visionplastic contaminationcottonautomated inspection |
| spellingShingle | Mathew G. Pelletier John D. Wanjura Greg A. Holt Semi-Automated Training of AI Vision Models AgriEngineering machine vision plastic contamination cotton automated inspection |
| title | Semi-Automated Training of AI Vision Models |
| title_full | Semi-Automated Training of AI Vision Models |
| title_fullStr | Semi-Automated Training of AI Vision Models |
| title_full_unstemmed | Semi-Automated Training of AI Vision Models |
| title_short | Semi-Automated Training of AI Vision Models |
| title_sort | semi automated training of ai vision models |
| topic | machine vision plastic contamination cotton automated inspection |
| url | https://www.mdpi.com/2624-7402/7/7/225 |
| work_keys_str_mv | AT mathewgpelletier semiautomatedtrainingofaivisionmodels AT johndwanjura semiautomatedtrainingofaivisionmodels AT gregaholt semiautomatedtrainingofaivisionmodels |