The spider monkey is an integral keystone species within the rainforest, dispersing seeds across large areas to propagate plants and trees, and help wildlife thrive. Due to deforestation and poaching they are now in danger of extinction and conservation is more necessary than ever before.
However, monitoring spider monkeys in the rainforest is incredibly difficult and time-consuming. Rangers, biologist and conservationists have to manually track their migration. Tracking was based on personal experience, which was inefficient and prone to human error. Can we use AI to recognize the screams of spider monkeys in order to achieve improved location tracking?
Using AI models to help detect spider monkeys can be extremely challenging within the rainforest though. The spider monkey call is very diverse in pitch and length, which is very different from the high-pitched scream we are familiar with. Most people are unfamiliar with these sounds, so being able to differentiate them requires a unique expertise.
The chainsaw model has provided similar technical challenges. The original model is used in a real-time rainforest environment. The mosquito sound, which is abundant within the environment, is often misreported as a chainsaw sound. Even a human ear has difficulty differentiating between the two sounds. Huawei needed to retrain the model using mosquito sounds and continuously adjust its accuracy.