AI in veterinary analysis methods
Reference number | |
Coordinator | Vidilab AB |
Funding from Vinnova | SEK 361 500 |
Project duration | October 2019 - October 2020 |
Status | Completed |
Venture | AI - Competence, ability and application |
Call | Start your AI journey! |
Purpose and goal
The purpose of the project was to develop an AI which can classify parasites from animals. Through this the AI competence within the company will also increase. During the project an AI has been developed and the companys AI skill has grown by experiences from challanges in annotation of data, training and validation. The classifier is well on it´s way to reach production quality. With the newfound knowledge in AI this quality is expected to be reached within the comming months.
Expected results and effects
The result was an expanded knowledge and experiences in pitfalls and challanges in AI. This knowledge will help Vidilab to identify good future projects for AI. A fast AI prototype for classification of horses with close to 99% accuracy across the board was also developed. This prototype will be refined and is expected to launch during next year. The prototype is expected to help Vidilab expand both nationally and internationally the following years.
Planned approach and implementation
Over 100 samples from production was scanned by an automatic microscope. The samples were annotated by experts to close to 3000 objects. Two AI plattforms were tested for training of the networks. A plattform for optimizing hyperparameters was also developed. Part of the data that was excluded from all training was used for validation.