Project Title BioSound: Deep Learning for Bird Sound Classification in the Western Ghats
Problem Statement Monitoring bird populations is essential for tracking ecosystem health and biodiversity. However, traditional bird surveys are costly, time-consuming, and limited in scope—especially in vast, ecologically sensitive regions like the Western Ghats.
AI Solution Overview BioSound uses passive acoustic monitoring and deep learning to automatically identify bird species from audio recordings. The system preprocesses sound data by converting it into mel spectrograms, optionally applying noise reduction to improve clarity. A convolutional neural network (CNN) trained on labeled bird call datasets then classifies the audio into one of 180+ species. The model is deployed via a user-friendly Streamlit interface that displays predictions, confidence levels, and spectrograms. This approach allows scalable, high-resolution biodiversity monitoring without the need for constant human presence.
Impact & Ethics By automating biodiversity assessment, AvianAI supports conservationists in identifying ecological trends, tracking restoration success, and protecting endangered species in biodiversity hotspots. Ethical considerations include preventing misuse (e.g., poaching), ensuring transparency in AI predictions, and respecting local ownership of ecological data.
Technologies Used Languages: Python Libraries: TensorFlow / Keras, KerasCV, NumPy, Librosa, Noisereduce, PIL Frameworks: Streamlit AI Models: Custom-trained CNN for audio spectrogram classification Data: Curated bird call dataset (150 species from India)
What We Learned / Challenges Faced We learned the importance of careful audio preprocessing—especially when working with faint or noisy wildlife recordings. One key challenge was ensuring that noise reduction did not eliminate important bird vocalizations. We also faced limitations due to the lack of high-quality labeled datasets, which we addressed through augmentation and model fine-tuning. Overall, we gained valuable insights into real-world AI deployment in conservation.
Log in or sign up for Devpost to join the conversation.