Machine Learning Project

Overview

This project explores the BirdCLEF 2026 dataset and aims to predict, from audio recordings, the probability of presence for each bird class.

Dataset

• Competition: https://www.kaggle.com/competitions/birdclef-2026

Project Constraints

• No neural networks (NN)

Goal

Predict from sound the probability of presence of each bird class.

Running the code

Quick start: Run training from config

A simple script is provided to run training from a YAML configuration and save artifacts (predictions and models).

1. Edit the config file at configs/train.yaml to adjust paths and parameters.
2. Run the training script:

python scripts/run_train.py configs/train.yaml

Outputs will be saved to the predictions/ directory and per-class models to models/ by default.

Authors

• Chu Amélie
• Tambellini Alan