Chatak

We turn massive, hard-to-analyze wildlife audio into instant, clear biodiversity insights—affordable for NGOs, effortless for researchers, and simple enough for any community to use.

Comment

Project Story — Chatak Eco Soundscape

About the Project

Chatak Eco Soundscape was created to make biodiversity audio analysis faster, accessible, and practical. Researchers and conservation teams collect thousands of hours of field recordings but lack simple, affordable tools to process them. This project fills that gap with an AI-powered, all-in-one acoustic analysis platform.

What Inspired Us

We saw scientists spending weeks manually reviewing audio, while existing tools were too expensive, fragmented, or required coding knowledge. Many NGOs and field teams simply couldn’t use advanced bioacoustic tools because they weren’t built for them. That inspired us to create a platform that democratizes acoustic biodiversity monitoring.

What We Learned

  • Acoustic analysis has multiple layers: segmentation, feature extraction, detection, clustering, annotation.
  • Location-aware AI models greatly improve species detection accuracy.
  • Offline-first tools are crucial for remote fieldwork.
  • Users prioritize speed, clarity, and simple workflows.
  • Transparent AI outputs help build trust with ecological users.

How We Built It

  • Developed an offline-ready desktop app using Electron + React.
  • Integrated BirdNet and PANNs for species-specific and general acoustic event detection.
  • Built a structured pipeline for segmentation, inference, clustering, and annotation.
  • Implemented HDBSCAN for discovering patterns and unknown sound groups.
  • Added spectrogram views, snippet extraction, and expert validation interfaces.
  • Used PostgreSQL with local/S3 storage for scalable data handling.
  • Ensured reproducibility through isolated Python environments.

Challenges We Faced

  • Processing long audio files efficiently without memory issues.
  • Handling model inaccuracies in noisy, real-world recordings.
  • Maintaining performance while running heavy AI workloads.
  • Designing a simple UI for a complex backend pipeline.
  • Dependency issues with Python, FFmpeg, and ML libraries.
  • Merging multiple AI engines into a cohesive workflow.

Built With

Share this project:

Updates