Bird Classification using CNN

Our motive is to build a Bird classifier using CNN classification.This project would help people to identify birds in real time-made possible using opencv videoCapture.

Comment

Inspiration

We wanted to build a model that will help people in bird watching, because sometimes if they want to get information about an unknown species of a bird, they can just take the images of the bird and run through our system to get information

What it does

It takes the images of the bird, run through our CNN model and predicts the species of the birds, it can also make a prediction using OpenCV on the live video feed.

How we built it

We used a dataset from kaggle and use CNN, we made 5 hidden layers with relu activation function and Softmax for output layer. We achieved an accuracy of around 90% and that model file we integrated in our webapp using flask. We also made video.py file where the model file will be loaded and prediction will be made on live video feed from the camera.

Challenges we ran into

The dataset was big having around 300 classes with over 50,000 thousand images and the estimated time to train was around 18-20 hours, so we reduced the classes to 150 training our model under 8 hours. We couldn't implement live video feed feature on webapp , so we made a separate file for that feature.

Accomplishments that we're proud of

Achiving an accuracy of 90%, completed the project before deadline

What we learned

CNN, flask

What's next for Bird Classification using CNN

implementing live video feature on webapp or using REST API and react, increasing accuracy and training 300 classes

Share this project:

Updates