Bird Recognition App | Businessware Technologies

Case Studies

Bird Recognition App

Technologies:
Industry:
Webcam Services
Client:
Platform:
Mobile, Desktop
Country:
USA
99%
Accuracy
>120
Bird species
Bird Recognition App

Project Summary

An AI-based bird detection system for backyard wildlife observation. Real-time detection and observation of over 120 bird species and 20 mammal species

Services

MVP development
Dataset management

Team

1 Project manager
3 Machine learning developers

Target Audience

Birdwatching & Wildlife Protection

Case Study

We were approached to create a service for those interested in observing wildlife - a video camera service with real-time bird detection coupled with a mobile app. The user installs an IoT camera in their backyard near a feeder, and as soon as a bird enters the frame, the user receives a notification including a bird's species along with a video recording.

Challenge

We had to work within the constraints of the IoT cameras as they are not capable of running heavy object detection neural networks typically used for a project like this, while maintaining a high degree of detection accuracy. The object detection module had to produce the least amount of false positives as possible (i.e. not get triggered by an insect flying by, a falling branch or a reflection in the water), as well as detect the object with as much accuracy as possible, providing accurate results to the user.

Another challenge was to accurately distinguish between different types of birds and animals. Many species of birds look very similar, making it hard for a neural network to differentiate them. When creating our recognition model we had to take into account the subtle difference between species to provide the most accurate result. Moreover, young birds often look much different from the adult ones, so we had to work on correctly detecting both adult and baby birds.

Solution

The service we have developed is capable of detecting over 120 species of birds in real time with a high degree of accuracy. The neural network runs directly on IoT cameras, detecting animals which enter into frame, triggering video recording and generating thumbnails to provide content for the users.

Each video includes tags which describe the species of the bird. Users can send inappropriately tagged videos for analysis to improve detection or tag the videos correctly themselves. The system is capable of accurately detecting both adult and baby birds, as well as determining their sex. The neural network can also detect other animals which can enter the frame, like raccoons.

Results

The solution we have worked on has been successfully implemented and released on the market. The animal recognition system has already gained traction and garnered the interests of birdwatchers and animal enthusiasts all around the globe, including celebrities.

Let's Work Together!

Do you want to know the total cost of development and realization of the project? Tell us about your requirements, our specialists will contact you as soon as possible.