Limitations, Next Steps, and the Road Ahead

VScreen is a working research direction. Its next phase is about stronger evidence, broader coverage, and safer deployment.

Current limitations

Acoustic feature extraction can be computationally expensive, especially on constrained devices, and requires careful signal-processing choices.

Audio features and spectrogram features do not always align cleanly, making it difficult to combine every useful representation in one model.

Recordings from one research setting cannot represent every breed, farm layout, microphone, climate, or background-noise profile.

Disease identification is nuanced. A model output should prompt investigation and veterinary review rather than stand alone as a diagnosis.

Add complementary datasets from other geographical regions and evaluate performance across varied farm environments.

Develop and test a complementary model that can interpret the combined acoustic pattern of a flock.

Extend the Android and web workflows so an alert can support timely communication and professional follow-up.

Integrate IoT and robotics, improve edge efficiency, and investigate applications beyond poultry where the evidence supports them.