AndryRafam/Fruit

Fruit Recognition

Fruit Recognition

This little project is about classifying fruits (33 different types). In order to achieve our goal, we are going to use two methods:

- 1) Building our own model from scratch.
- 2) Using Transfer learning techniques. 

Dataset: https://www.kaggle.com/sshikamaru/fruit-recognition

Experiment Result

Model from Scratch:

- Validation Accuracy: 99.73 %
- Validation Loss: 0.85 %

VGG16 - Transfer Learning (Fine Tuning):

- Validation Accuracy: 99.85 %
- Validation Loss: 1.61 %

MobileNetV2 - Transfer Learning (Feature Extraction):

- Validation Accuracy: 100.00 %
- Validation Loss: 2.96 %

ACKNOWLEDGMENT

Thanks to Edward Zhang (https://www.kaggle.com/sshikamaru) for this dataset.