Bird Species Identification with Deep Learning

Tags: Deep Learning | Spectrogram | Audio Analysis

Link: GitHub

🔍 Overview

Trained a neural network to classify birds by their calls using spectrogram images from Xeno-Canto.

❓ Key Questions

• Can deep learning identify species from audio patterns?

• Which species are most often misclassified?

🧪 Methodology

• Spectrogram preprocessing

• Used Keras Sequential with dropout and softmax layers

📈 Results

• Accuracy: 95.45% (binary) and 67.24% (multi-class)

• Most confusion between Blue Jay and Western Meadowlark