Classification of fruit images using techniques such as the transfer learning, pruning, model trimming and custom model development. An attempt to adapt the model to multi-label classification with real-life examples.
Konrad Pawlik - konradpawlik@student.agh.edu.pl
Jan Fiszer - fiszer@student.agh.edu.pl
A thorough report detailing the project, the models selected and the decisions taken can be found in AML-Fruits360-Report.pdf
- Python3
- numpy
- matplotlib
- pandas
- tensorflow
- tensorflow_model_optimization
- exploratory_data_analysis.ipynb
- transfer_learning.ipynb
- custom_model.ipynb
- multi_label_and_pruning.ipynb
- data
- fruits-360_dataset
- fruits-360
- Training
- Test
- test-multiple_fruits
- fruits-360
- fruits-360_dataset
The original dataset contains images of the fruit in the native resolution, but these have been omitted due to their scarcity.
Data can be downloaded from https://www.kaggle.com/datasets/moltean/fruits?datasetId=5857&sortBy=voteCount