A deep learning-based classification method for palm fruit maturity using CNN models, including ResNet50 and InceptionV3. This project leverages transfer learning with fine-tuning to classify palm fruit bunches into five maturity stages: unripe, under-ripe, medium-ripe, ripe, and over-ripe.
- Deep CNN models: Baseline CNN, ResNet50, and InceptionV3
- Transfer learning with pre-trained models
- Dataset preprocessing and augmentation
- Evaluation metrics: Accuracy, AUC, F1-score, ROC curves, and confusion matrices
The original dataset can be found here and was modified for this study. The modified version, which can be downloaded here, consists of images captured in natural orchard environments. The dataset was split into 80% for training and 20% for testing.