Small project to identify photos with and without birds using a convolutional neural network (CNN) from TensorFlow (Keras library).
This project was build using Google colab. The notebook Birds.ipynb can also be accessed on the link:
https://colab.research.google.com/drive/1MulvuK2NKyivPijppWD6BjofQjwHxFgS
The images used to train the CNN are not available due to space. However, the weights of the trained model are saved in the file checkpoint, that can be recovered using model.load_weights(checkpoint_path).
In the last epoch trained, the perfromance of the model was: ~97% accuracy on the traning set (70% of the images) and ~95% on the validation set (30% of the images).
The script bird_identification.py was used to label the original images, moving them to their respective folder ('birds' or 'nobirds').