Overview / Usage

In recent years, birds have been losing their habitat due to pollution and diseases. It is imperative to monitor their health to prevent their extinction. Consequently, bird species identification is a preliminary step to tackle such a pressing problem.

Methodology / Approach

1- Explore data (number of records, classes, etc.)

2- Prepare data (detect outliers, extract features, split data into train and test sets, etc.)

3- Fit data into a deep-learning model

4- Evaluate model

Steps 2 and 3 can be repeated (extracting features, fine-tuning model) until reaching a good performance

​

​

Technologies Used

Pillow: image processing

Librosa: signal processing

Tensorflow: deep learning

Vision Transformer (ViT): a pre-trained computer vision model

Repository

https://github.com/Sinda271/BirdSpeciesClassification