Animal Detection using CNN

Our project is designed to help identify animals using Convolutional Neural Network(CNN) classification and to serve as an alert system for people in their practices.

Comment

Inspiration

To Help People to surveillance the animals in their homes or even farmers to protect their farms.

What it does

It is based on CNN where the user uploads the images on the web app and the prediction will be made, if the predicted class is an animal an SMS will be sent to the person via Twilio on the mobile number he entered in the form. There is also a video.py file that uses OpenCV and does real-time prediction.

How we built it

We used Convolutional Neural Network, used 5 hidden layers with relu activation function and softmax with output layer. The accuracy is over 90%. This model file we integrated in our web-app using flask and implemented twilio API where is the class human is not predicted, an SMS is sent that the creature in the camera is an animal.

Challenges we ran into

We had to reduce the dataset because the original dataset would estimated take 6-7 hours to train, this one took around 2-3 hours. Also we couldn't implement live video feature in our web app, that's why we made a separate file for demonstrating that feature.

Accomplishments that we're proud of

Achived an accuracy of over 90%, successfully implemented twilio API

What we learned

Flask, git and github to work in teams, CNN

What's next for Animal Detection using CNN

Implementing live video feed on web-app or implement using REST API and react, increasing accuracy

Share this project:

Updates