Inspiration
The inspiration behind this system is the need to promote coexistence between wildlife and humans in urban areas. As urbanization continues to encroach on natural habitats, there is a growing need to manage wildlife populations and minimize conflicts between humans and animals. The automated wildlife tracking system will enable more efficient and accurate monitoring of animal populations, and help NParks and SPCA to implement effective population control and relocation programs.
What it does
An example would be the recent stray dog problem in which stray dogs have been attacking community cats in area like Ang Mo Kio. Our system would allow us to identify groups of such stray dog population through CCTV footage and alert SPCA. With this system, SPCA can efficiently monitor and track the population and behaviour of such stray dog 'gangs' and take preemptive measures before attacks happen. Animal populations tracked by our system will be recorded and displayed on a user-friendly dashboard and will quickly convey data insights required by relevant agencies.
How we built it
We obtained training data from the Open Image dataset, training it on ultralytics_v8 YOLO model. The model is deployed on Amazon Sage maker and will be used to overlay CCTV footage. The data obtained will be stored in the database upon which we use Tableau to build a user-friendly dashboard for relevant agencies
Challenges we ran into
Some challenges that we run into were deploying and writing code using the AWS console. Due to the AWS console being a foreign environment for all of us, it took additional time and effort to be familiarize with and work in the conditions provided.
Accomplishments that we're proud of
We are proud to have successfully trained a model that is able to identify different animals on a live video footage. We are proud at the speed in which we learn to work with a foreign environment in which none of us has had the exposure to.
What we learned
Throughout the development process of the automated wildlife tracking system, we have gained valuable knowledge and experience in several areas. We learned how to train a YOLO model using the Open Image dataset and deploy it on Amazon Sage maker. We also learned how to use Tableau to build a user-friendly dashboard to display animal population data insights. Additionally, we gained experience in working with AWS console and overcoming challenges that came with it. This project allowed us to develop our technical skills and apply them to a real-world problem, while also learning to collaborate effectively as a team.
What's next for Watchdog.AI
To improve and upscale the automated wildlife tracking system for the future, several steps can be taken. Integration with AI and machine learning technologies can enable advanced data analysis and predictions based on the animal population data collected. Additionally, incorporating a wider range of data sources such as satellite imagery and drones can increase the coverage and accuracy of the tracking system. Collaborating with other organizations and agencies to share data and insights can facilitate a more comprehensive and coordinated approach to wildlife management. Developing mobile applications that enable the public to report animal sightings and incidents can facilitate a more collaborative and community-based approach to wildlife management. Continuous evaluation and improvement of the system to ensure that it remains up to date with the latest technologies and best practices in wildlife management are also necessary. By taking these steps, the automated wildlife tracking system can be improved, scaled up and made more effective in promoting coexistence between wildlife and humans in urban areas.
Log in or sign up for Devpost to join the conversation.