University of Twente Student Theses
Assessing the Impact of Including Diseased Crop Images in Training Datasets on the Performance of Convolutional Neural Networks for Crop Classification
Lelasseux, Quincy (2025) Assessing the Impact of Including Diseased Crop Images in Training Datasets on the Performance of Convolutional Neural Networks for Crop Classification.
This is the latest version of this item.
|
PDF
2MB |
| Abstract: | This study investigates the impact of including diseased crop images in training datasets on the performance of ResNet18, a convolutional neural network(CNN),incropclassification tasks. The focus lies on assessing model performance in scenarios where both healthy and diseased crop images are present in the dataset. Under these conditions, the model can adapt and classify crops more effectively by fine-tuning the CNN using transfer learning. Datasets containing varying proportions of diseased crop images (e.g., 0%, 10%, 30%, and 50%) were systematically analyzed to evaluate the effects of these variations on classification accuracy, precision, and F1-score. Experiments included training with both mixed and separated datasets as baselines for discussion. To ensure robust and reliable results, the training and evaluation process incorporated k-fold cross-validation. In the results of this study it was found that the absence of diseased samples during training significantly reduces the model’s ability to generalize to real-world conditions, whereas incorporating such images enhances robustness and accuracy. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/105107 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page