Model Transfer Learning EfficientNetB0

About this project

Implemented a CNN-based bird species classifier using EfficientNetB0 transfer learning to distinguish four grain‑eating bird species and an unknown class from images collected via eBird.org. The model, built in Python with TensorFlow/Keras, was trained on 224×224 RGB images, evaluated with stratified 5‑fold cross‑validation, and achieved ~82 % test accuracy. The project leverages Jupyter Notebooks, Google Colab GPU runtime, and libraries such as NumPy, pandas, OpenCV, and scikit‑learn for data preprocessing, training, and performance reporting.

Links