Fetchr

Inspiration

We were driven by a basic yet common concern shared by pet owners: is my pet safe when I'm not there? Be it a dog running off, a cat behaving unusually, or a potential medical emergency, there is no way of knowing right away. We wanted to change that. Inspired by wearable health trackers for humans, we sought to build a smart collar that gives pets a voice, a way of safety, visibility, and peace of mind with real-time data and smart alerts. That is precisely what Fetchr is: a system to help owners understand what their pets are doing and if they are okay anytime, anywhere.

What it does

Fetchr is an real-time pet safety system using hardware, software, and AI to track a pet's movements, location, and overall behavior. A light-weight collar containing an ESP32 microcontroller, accelerometer/gyroscope, and a GPS allows for ongoing tracking of a pets activity and position. The data is sent to Firebase in real-time and then mapped and visualized in a SwiftUI mobile application. Live maps and activity dashboards allow owners to see where their pet is located and how active they are. Additionally, a machine learning model analyzes the real-time data being gathered from the sensors to identify abnormal behaviors such as rapid falls, excessive activity, or moving outside the safe zone boundaries, and send alerts to the owner immediately. All in all, Fetchr is a smartwatch and safety system for pets, providing families with peace of mind while attempting to prevent emergencies.

How we built it

The initial phase was focused on creating the physical part of the system utilizing an ESP32 microcontroller connected to a gyroscope, accelerometer, and GPS module. The ESP32 collects data from the sensors and sends it via Wi-Fi to a Firebase Realtime Database, which allows for dynamic updates with low latency and no data loss. For the software side, we developed an iOS app using SwiftUI that fetches the real-time data and displays it on a live map along with comprehensive activity statistics and analytics in chart and graph formats. For the AI aspect of the project, we recorded the motion data provided by the sensors and trained a lightweight machine learning model to differentiate between normal and abnormal patterns of movement. The final challenge was integrating math parts, cloud portions, the app, and AI, but in doing so, we built a complete, fully functioning solution that works seamlessly from end to, end.

Challenges we ran into

Developing Fetchr was a challenging endeavor that involved resolving many technical and design issues. One of the biggest hurdles was achieving real-time hardware synchronization with Firebase and the mobile app without additional latency or dropping data. Furthermore, we had trouble working with noisy sensor data that often resulted in confusion distinguishing between natural movement and legitimate anomalies. Power management for the ESP32 was also difficult since running GPS and Wi-Fi would quickly drain the battery. Lastly, we needed to optimize the machine learning model such that it would work with limited resources while retaining a high level of prediction accuracy. Because we had a set amount of time to complete everything during the hackathon, managing all of these components tested our teamwork and problem-solving abilities, and improved the quality of the project.

Accomplishments that we're proud of

We take pride in that we were able to combine hardware, AI, and software into a single working product through the hackathon event. We created a real-time tracking system that streams and displays the movement data in a user-friendly mobile interface. Our machine learning model was able to detect unusual activity with impressive accuracy, and the app provided real-time alerts when it did. But beyond the technical accomplishment, we are most proud that Fetchr clearly shows how technology can bring real and meaningful value to people's lives to provide safety for their pets and more peace of mind.

What we learned

During this project, we discovered a lot about integrating hardware and software in real-time, as well as how to manage sensor data for machine learning. We learned to experience IoT communication, data processing, and cloud synchronization. Working across disciplines pushed us to consider the big picture involving engineering, design, and data science to solve real-world problems. We also learned that designing with the user in mind is so crucial, making sure the system is not only technically sound, but is also intuitive and reliable for everyday pet owners.

What's next for Fetchr

In the future, we plan to enhance Fetchr to be smarter, smaller, and more functional. We expect to add health-tracking sensors such as heart rate and temperature monitors next to provide greater insights into a pet’s health status. We also want to expand the AI model to include geofencing and route learning capabilities so Fetchr can detect when a pet is off its normal path. Capabilities for offline alerts via Bluetooth, support for Android, and an enhanced web dashboard are also development agenda items. Over the longer term, we see Fetchr evolving into a full pet safety ecosystem that would not only help pet owners, but could also be useful for shelters, veterinarians, and animal rescues to identify early signs of health or behavior issues in pets.