SpotZoo

SpotZoo:

Live and Let Live - Using AI and Crowdsourcing to stop unexpected animal attacks.

SpotZoo is a community-driven ML-powered mobile application that allows users to crowdsource real-time information on animal encounters so that other people in that location can avoid a possible animal attack. Through our app, users can provide photos of animals they saw or encountered & add them to a geographic database that other users can see. We made a mobile app where users can log in, submit photos (&/or even click them) and then it's deployed to IPFS which later on is fed into an ML model running in the backend to leverage Computer Vision which helps in recognizing the object which helps to identify the animal & add the encounter to Firebase realtime database. We also utilize Google maps APIs to provide an interactive, location-based viewer for encounters and alerts when you might be getting close.

Animals ? are everywhere, right? Even in Massachusetts, where we have everything from snakes to Coyotes to Bobcats. Though some animals are a man's best friend, like your very own dog, some are the best enemies as well. Humans and nature are always in conflict.

Now, the mess gets worse. Humans are gradually snatching the shelter and the habitat of animals and animals are forced to live closer and closer to humans. Due to this increasing proximity, several unwanted and unexpected encounters between humans and the wild have started occurring frequently. As a consequence, Every year, people die due to these unexpected and dangerous animal attacks. Even the animal is injured or killed for the crime of “killing a human”. In rural areas and third-world countries, it's considered common to die due to animal attacks.

Thousands die and millions are left severely injured worldwide by animal attacks. Injuries, too, cost a lot to treat and manage, and in the US alone, treatment of animal injuries costs over $2 billion.

We propose a noble solution to this problem. A community-driven highly secured mobile application that leverages the cutting-edge power of ML & allows crowdsourcing real-time information on animal encounters so that other people in that location can avoid a possible animal attack.

Thousands die and millions are left severely injured worldwide by animal attacks. Injuries, too, cost a lot to treat and manage, and in the US alone, treatment of animal injuries costs over $2 billion. Even the animal is injured or killed for the crime of “killing a human”. In rural areas and third-world countries, it's considered common to die due to animal attacks, For example, every day in India, on average twenty-eight people die due to an unexpected wild animal attack. This was an alarming big number for us to work upon.

Most of our targeted audience are people living in suburban/rural areas or people who have a frequent danger of animal attacks or who live near wildlife habitats. They can download the app from the play store then, After successfully logging in, the app requests you to share your current location. While reporting an animal you upload an image of that animal and the app immediately recognizes the animal. You add more info's like, time spotted, number of animals etc. After submitting the location of the animal is posted on the interactive map along with the spotting info. All other users within a 5km radius are notified about the animal, they can see its picture and the time spotted. Henceforth, they will get alerted and take precautions and not go near the spotted place.

We have conducted various surveys gathered over 1000 responses, got multiple reviews and feedbacks. Got in contact with many people from targeted users. Reviewed many opinions and changed and reconsidered many features. Emphasized security and authenticity of the application.

Animals ? are everywhere, right? Even in Massachusetts, where we have everything from snakes to Coyotes to Bobcats. Though some animals are a man's best friend, like your very own dog, some are the best enemies as well. Humans and nature are always in conflict.

Now, the mess gets worse. Humans are gradually snatching the shelter and the habitat of animals and animals are forced to live closer and closer to humans. Due to this increasing proximity, several unwanted and unexpected encounters between humans and the wild have started occurring frequently. As a consequence, Every year, people die due to these unexpected and dangerous animal attacks. Even the animal is injured or killed for the crime of “killing a human”. In rural areas and third-world countries, it's considered common to die due to animal attacks.

Thousands die and millions are left severely injured worldwide by animal attacks. Injuries, too, cost a lot to treat and manage, and in the US alone, treatment of animal injuries costs over $2 billion.

We propose a noble solution to this problem. A community-driven highly secured mobile application that leverages the cutting-edge power of ML & allows crowdsourcing real-time information on animal encounters so that other people in that location can avoid a possible animal attack.

We currently have a developed prototype for the app. We also submitted to Penn Apps (University of Pennsylvania's hackathon where the app won 4th place) - View it here - https://devpost.com/software/s...

The app is pretty complex and took lots of effortful hours to make it. Let's divide the app into 3 categories:

1. Frontend - The app was designed in Figma, and then the frontend was coded in Flutter. We use Google Maps API to fetch users' locations and show alerts within a 5km radius. The user authentication was done with Firebase.

2. ML Model - We used MobileNet V2 SSD to build our Machine learning model. We gathered over 1000+ images approximately 2 GB of data and the model was trained 150,000 times. After this, we reached an accuracy of 92%. The ML model was then integrated into the backend.

3. Backend - We decided to proceed with firebase/firestore as our database and had an independent node js/express backend running on the Google App Engine.

4. We used IPFS to protect the mutability of the image and make it more secure and easy to transfer. We had an ML model to verify that the image is an animal only and enhance authenticity of the app.
   

It just exists as a prototype right now!

We plan to develop more and launch the fully-fledged seamless app on the play store and app store. Contact government for appropriate help and advertise the product on a huge scale to get more targeted users to use it!

We have successfully implemented an application that can help people survive in devastating situations of encountering animals and allow people to discover animals around them at the same time. Moreover, we have orchestrated high-end technologies including Cloud, IPFS, and a Machine Learning model to thoroughly build our application. Consequently, we are proud of finishing the project on time which seemed like a tough task as we started working on it quite late due to other commitments and were also able to add most of the features that we envisioned for the app during ideation. And as always, overnight working was pretty fun! :)

A lot! Initially, we were facing a problem setting up the TensorFlow model on our project as it was a very buffed one and flutter doesn't accept this format so we had to reduce the dataset parameters & then optimize it so that it can run seamlessly with low latency. Then creating an API endpoint was also a big issue.

We have the prototype and demo ready.

Zero Yet