Plant Classification in Ios: Integration of Machine Learning Models Through Core Ml
Session
Computer Science and Communication Engineering
Description
This paper addresses the integration of pre-trained Machine Learning models into iOS applications through the use of Core ML. The primary objective was the development of an application for plant classification by employing an existing model trained on the PlantCLEF 2024 dataset. Initially, an analysis was conducted on available models and the challenges encountered during the conversion of a PyTorch (.pth.tar) model into the .mlpackage format suitable for iOS. The process involved input normalization, model tracing, the addition of a Softmax layer, and the configuration of necessary parameters to preserve the accuracy of the converted model. Once the model was integrated into the application, testing was carried out using real-world images and data from PlantCLEF, with results compared to those obtained from the original PyTorch model. The findings demonstrated satisfactory alignment in classification, although the iOS application exhibited a lower confidence level compared to the source model. This indicates that while conversion does not significantly affect classification accuracy, it may influence the model’s confidence output. Through this project, in-depth insights were gained into the practical challenges of working with ML models and adapting them for mobile deployment.
Keywords:
Plant classification, iOS, Core ML, Machine Learning, PyTorch conversion, Mobile applications, PlantCLEF
Proceedings Editor
Edmond Hajrizi
ISBN
978-9951-982-41-2
Location
UBT Kampus, Lipjan
Start Date
25-10-2025 9:00 AM
End Date
26-10-2025 6:00 PM
DOI
10.33107/ubt-ic.2025.82
Recommended Citation
Prestreshi, Rina; Qehaja, Besnik; and Hajrizi, Edmond, "Plant Classification in Ios: Integration of Machine Learning Models Through Core Ml" (2025). UBT International Conference. 14.
https://knowledgecenter.ubt-uni.net/conference/2025UBTIC/CS/14
Plant Classification in Ios: Integration of Machine Learning Models Through Core Ml
UBT Kampus, Lipjan
This paper addresses the integration of pre-trained Machine Learning models into iOS applications through the use of Core ML. The primary objective was the development of an application for plant classification by employing an existing model trained on the PlantCLEF 2024 dataset. Initially, an analysis was conducted on available models and the challenges encountered during the conversion of a PyTorch (.pth.tar) model into the .mlpackage format suitable for iOS. The process involved input normalization, model tracing, the addition of a Softmax layer, and the configuration of necessary parameters to preserve the accuracy of the converted model. Once the model was integrated into the application, testing was carried out using real-world images and data from PlantCLEF, with results compared to those obtained from the original PyTorch model. The findings demonstrated satisfactory alignment in classification, although the iOS application exhibited a lower confidence level compared to the source model. This indicates that while conversion does not significantly affect classification accuracy, it may influence the model’s confidence output. Through this project, in-depth insights were gained into the practical challenges of working with ML models and adapting them for mobile deployment.