Overview

The Doodle classifier is based on a Convolutional Neural Network which classifies the doodle input given by the user in 20 different classes.

​
Abstract:
  • Computers are advancing at great speed in this age.
  • If they have the ability to understand our doodles or quick line drawings, it will allow for much more advanced and simplified forms of communication.
  • Drawing and understanding images is a way of communication when words fall short or language is inadequate due to different cultural and literacy levels .
  • The techniques and algorithms used in training a machine to do specific tasks are grouped under a subject called Machine Learning and Neural Networks.

Approach:

​
  • The preliminary stage involved studying and learning the basics of Machine Learning and Deep Learning algorithms.
  •  
  • For the better understanding of the topic, we first developed a Digit Classifier from scratch using the MNIST dataset using Numpy library.
  • All the functions were build from scratch for the Forward as well as Backward propagation.
  • The CNN model is build with the help of PyTorch library for the convolution of image with filters along with maxpooling.
  • After multiple convolutional layers, the input representation is flattened into a feature vector and passed through a dense neural network for the output. A Drawing Pad is created using OpenCv for getting input from the user.
  • PySimpleGUI is used for implementing GUI.

Model Architecture :
Convolutional Connected Layers:
Layer
​Kernel size
Filters
​Maxpool
​Padding
Conv1
​(5,5)
8
​(5,5)
1
Conv2
(5,5)
16
(5,5)
1
Conv3
(3,3)
28
​None
1
Conv4
(3,3)
​28
None
1
Fully Connected Layer:
Layer
Size
Fc layer1
(48x5x5,500)
Fc layer2
(500,250)
Fc layer3
(250,20)
Hyperparameters:
​Parameters
Values
Learning rate
0.01
Epochs
100
Batch size
1200
Beta
0.9
​Optimizer
SGD
​Loss function
BCE Loss
Result
Dataset
​Accuracy
​Training dataset
​ 95.77 %
Testing dataset
95.74 %
GitHub Repository

Tools And Libraries used

​Team Members:
  • Kishan Verma
  • Rutu Shrirame
  • Yash Gaikwad
  • Shyam Lakhan Sah
  • Jaswanth Karanam
  • Srivardhini Dasarapu
  • Riya Deshpande
​Team Mentors:
  • Tanmay Pathrabe
  • Diksha Bagade
  • Muhammed Abdullah
  • Siddharth Singh
  • Aneesh Shetye
  • Kshitij Ambilduke
  • Akshansh Malviya
  • Pulkit Mathur
  • Sibam Parida