Inspiration
The recent Kerala news headlines are surged with news of human wildlife conflicts and the resultant loss of life and livelihood. Nearly 30 percent of Kerala's geographical area is forested and nearly 30 lakh people are living on its peripheries. According to the latest data over 640 lives were claimed in human wildlife conflict in the last 5 years in Kerala. This is alarming and many of these lives are lost due lack of proper warning messages to the public.
We came up with ‘WildCatch’ , a humble solution to this hot issue.’WildCatch’ is a FOSS tool that provides real time warning messages to the public living close to forest.This software detects the presence of wild animals near human settlements from cameras installed there, and gives real time warning messages to the locals.
What it does
It primarily solves the problem of time delay of communication of danger to the public by automating the whole process. It helps the locals to take precautionary measures as early as possible. The solution can also be extended in use for the tourists in wild areas. The other main attractive point of the solution is its capability in giving customized warning messages based on the type of animal identified. The warning message consists of “what to do '' based on the animal ,as each animal behaves differently on arriving at human settlements. The tool also gives primary measures to deal with that particular animal.
How we built it
This tool uses the ML and the wide spread wildlife surveillance infrastructure of the forest department. These cameras record footage of wild animals. Our model uses these footages and identifies the presence of wild animals in these areas. We'll be building the model by training the YOLO V8 model by our custom dataset of wild animals. We'll be implementing it using OpenCV. YOLO ("You only look once") is an open-source image analysis AI system developed by the computer vision community. If the surveillance camera visuals contain the presence of wild animals , it will be detected by our model, which in turn will give out an alert to all people living nearby to where the presence was found.The messages are sent via mass messaging method. We'll be using SimpleTexting, whatsappJs for the mass messaging of alerts. The custom messages are made based on the behavior of animals. The location is identified from the source of visuals .
Challenges we ran into
We had to collect large amount of data and labelling it was a bit of a task., Finding appropriate ML model, Integration of Whatsapp API, Live video acquisition from surveillance camera and transfer.
Accomplishments that we're proud of
It accomplishes the task of saving hundreds of lives that may lose every during wildlife conflicts.
What we learned
Integration of API, Data labeling, Model training
What's next for WildCatch
We wish to partner with the forest department to implement the project in a larger scale., Increasing the accuracy of the model by incorporating larger pool of data.
Log in or sign up for Devpost to join the conversation.