UCS757:Building Innovative Systems

Submitted by:-

Rohan Dutt

101803151

COE 8

Projects

Live Link

https://ucs757-101803151-projects.herokuapp.com

Home

Project 3: Auto Stack Overflow Search Tool

Introduction

We all have faced issues while learning a new language or coding in the same. Well, this tool takes your python file and automatically figures out the errors and open up all the stack overflow pages which might help you.

Requirements

1. Python
2. Requests
3. NodeJS
4. EJS

Browser Requirements

Kindly allow popups to allow the app to function properly.

Live Link

https://ucs757-101803151-projects.herokuapp.com/auto-stackoverflow

Input Interface

Input

Output

Flowchart

Novelty

1. This is a new concept, of actually opening the webpages which might give the user the required solutions to the problems in their code
2. Piping the data from python, then receiving it in Nodejs and redirecting the user to stack overflow pages
3. The use case is quite common and will solve a lot of problems as we all face issues while learning a new language or working on a project.

Limitation

This tool is limited to python only for the time being and is dependent on the errors solutions to be present on Stack Overflow. Also, if the popups are not enabled, new webpages won't open.

Project 2: Age and Gender Detection

Introduction

The web app takes an image of a person as input and returns the image with the gender and age group details

Requirements

1. Python
2. OpenCV-Python
3. NodeJS
4. EJS

Live Link

https://ucs757-101803151-projects.herokuapp.com/age-gender

Input Interface

Input

Output

Flowchart

Observations

1. The age groups 0-2, 4-6, 8-13 and 25-32 are predicted with relatively high accuracy.
2. The output is heavily biased towards the age group 25-32 ( see the row belonging to the age group 25-32 ). This means that it is very easy for the network to get confused between the ages 15 to 43. So, even if the actual age is between 15-20 or 38-43, there is a high chance that the predicted age will be 25-32.
3. We observed that the accuracy of the models improved if we use padding around the detected face. This may be due to the fact that the input while training were standard face images and not closely cropped faces that we get after face detection.

Novelty

1. Detecting gender using an image, can also be extended to detect it in a video stream
2. Predicting the age group of all the individuals in an image
3. Using multiple build packs in Heroku for both Node.js and Python and experimenting with amazon ec2 for deployment

Limitation

This technique is limited with its efficiency in detecting age, it cannot detect exact age, and can only detect a certain age group. Also, unfortunately, it does not work on non binary genders yet.
One can also try to use a regression model instead of classification for Age Prediction if enough data is available.

Project 1: Mosaic Creator Using Node and Python

Introduction

Mosaic creator is a web app which takes the target image as input and generates that image using thousands of small images

Requirements

1. Python
2. NumPy
3. Pillow
4. NodeJS
5. EJS

Live Link

https://ucs757-101803151-projects.herokuapp.com/mosaic

Input Interface

Input

Output

Flowchart

Novelty

1. Generation of dataset for images(tiles), using extensions and small python scripts
2. Finding nearest neighbour in terms of colors for each predefined unit of the target image from the data set tiles
3. Randomization of images to ensure that maximum pixels are covered for each color
4. Using multiple build packs in Heroku for both Node.js and Python and experimenting with amazon ec2 for deployment

Limitation

This technique is dependent on the variety of images in the dataset in terms of colors, higher the number of images in each color, better will be the mosaic.