This project aims to classify 515 different bird species using ResNet, we will also explore model explainability through LIME.
https://www.kaggle.com/datasets/gpiosenka/100-bird-species
The dataset used for this project is obtained from Kaggle, containing images of 515 different bird species. The dataset is pre-split into training (82724 samples), validation (2575 samples), and test (2575 samples).
The CNN ResNet50 architecture was used to classify the bird species. The model was pre-trained on the ImageNet dataset and then fine-tuned on the bird species dataset.
After 20 epochs of trainning using CorssEntropy Loss and Adam Optimizer, we have:
- train loss: 0.0501
- train accuracy: 0.9848
- validation loss: 0.4280
- validation accuracy: 0.8971
