The problem Animal Recognisation solves

Due to decrease in forest lands, many times animals infilterate into villages, cities, highways and railway tracks which are present near or go through the forest. This endangers both, the people as well as the animals. Animals usully get killed by locals or are killed in automobile or train accidents. Also many times they destroy the farm lands, destroying the livelyhood of farmers.
We want to develop a system to protect animals from this unfortunate death due to the mistaken infiltration and farmers can use it to protect their fields from animals destroying their cultivation.
Our project generates a fear frequency for the animals so that they don't come near boundaries of such areas like highways or railway tracks. It is done by an early warning system which uses IR cameras to detect animals as soon as they come near the forest boundary.
We will build it using RaspberryPi with IR camera mounted, which senses the presence of animals. Using Machine Learning to detect the type of animal and generate fear frequency according to that type of animal making them return into the forest area. It uses decentralized system which is installed on trees at the boundary of the forest or any region which continuously monitor the track for any animal activity or suspicious activities like poaching and alert the nearest forest security force station.

Challenges we ran into

Challenges in our system were our sensors are limited to detect motions upto a limited range. Due to unavailability of high computing power GPUs and free cloud services we were unable to train our own model, henceforth we had to depend on APIs to serve our purpose, which provide limited free plans.
Since, our project involves various hardware components, reliablility of our project is always subject to good working condition of various IOT devices.