An e-commerce website facilitated with user login/logout, Facebook OAuth, Razorpay Payment API. Initially inspiration was to deploy both models to the website. But then only one ML model could be successfully deployed to website. ML model 1: Disease prediction based o symptoms ML model 2: Chatbot

Got a good job, rest you can see on link provided. (heroku baar baar un–deploy kr deta hai app bata dena re-deploy kar dunga)

Source 1 Source 2 Source 3

Link 1: (is course me course 3 app dev ka hai to can be skipped if not interested) There are three Udemy courses for ML 1. Pyhton bootcamp zero to hero 2. Python and R for Data Science 3. Machine Learning A to Z

Siddhartha Shukla

Modelled a path tracker bot which is capable of finding path to destination by optimising time and path length. Selected shortest path using Image Processing and controlled movement using Arduino. Tools and Technique used :- Machine learning, OpenCv, Arduino.

Explore machine learning field, and more about optimal path tracking concept to use in real life.

Coursera ML course, basic arduino, docs of OpenCv

Vishwajeet Kumar

We solved inverse kinematics for a legged robot using simple machine learning at first then using ANN.

Apply machine learning to a legged robot to get more efficient functioning.

Many sources available on the internet. Search for medium articles particularly.

Rashmi Deep Mahawar

You'll find everything in ppt

NA

Source 1

NA

Harsh Nadar

Automated Path Mapping Robot Sep '18- Oct '18 Constructed an autonomous robot that used an overhead camera and applied image processing techniques using MATLAB. Image Processing techiques were used to extract features from the image about the arena. The bot traversed the shortest path obtained from Dijkstra's Algorithm to perform some tasks.

NA

Just refer gfg for dijkstras also and basic image processing techniques(you can Google them)

Divya Gupta

It classifies the sentiments of a text review as either positive review or negative review. It used the knowlege of Tensorflow.

It gives the result that the accuracy of training set is more than that of validation set. It gives output as 0 or 1 depending on if the review is positive or negative.

Coursera

Niharika

Given a datasets , we have to apply ml algo to predict whether a person has cancer or not

NA

Source 1

NA

Pragati Pandey

The aim is to predict whether a person is COVID positive or not by examining the CT scan of the examinee.

NA

Source 1 Source 2 Source 3

Deep learning course: https://youtube.com/playlist?list=PLkDaE6sCZn6Ec-XTbcX1uRg2_u4xOEky0

Eshaan Gupta

* Developed a CNN based model using Pytorch that classifies the input CT scan images as Covid infected or as healthy ( not infected ). * Input images are filtered using a GRU Layer for better accuracy. * The accuracy obtained from the proposed CR Net model is 79% which is slightly lesser than that of some of the pre - trained standard models. * The prediction time of proposed model is lesser than that of pre - trained standard models. * The model is lighter and is suitable to use on the web.

NA

Source 1

https://archive.physionet.org/physiobank/database/

Rohit Yadav

Implementing Tiles algorithm for dynamic community detection and few other static community detection algorithms

This project is able to detect different clusters in the network and the results can be plotted for comparison

GitHub , Google

Swati

1. The task was to predict the level of diabetes of a patient by classification with the help of given data. 2. Data preprocessing and model training knowledge were a must to score well. I used a random forest classifier to train this model. 3. I used Flask to deploy the model on Heroku that was saved in a pickle file.

NA

Source 1

https://github.com/anshgoyal1/diabetes_prediction

Ansh Goyal

Detecting drowsiness of a driver from the facial expressions captured by a camera on the steering wheel, and thereby sounding an alarm

NA

Source 1

NA

Dinesh Easwaravaka

Categorised the action of the holder of the mobile into 6 different activities

Useable in tracking devices

Source 1

NA

Himanshu Balasamanta

1. Built a Human Emotion Detector which takes input from live video feed and detects the faces present at each instant and predicts emotions i.e. happiness, fear, sadness, disgust, anger, surprise and neutral. 2. Used the Kaggle FER-2013 dataset. 3. Built and trained a CNN model for emotion detection.

It detects the faces present at each instant and predicts emotions i.e. happiness, fear, sadness, disgust, anger, surprise and neutral.

Source 1

https://github.com/ashuKS29/Human-Emotion-Detector

Ashutosh Kumar Singh

Android app to detect type of activities the person is performing using accelerometer reading

NA

Source 1

NA

Sayam Rakshit

The project basically aims to identify suspicious electricity consumers by analysing their electricity consumption data of smart meters. These consumers have a wide irregularities in their electricity readings or very low consumption. Machine learning techniques can be used to visualise and analyse data.

The random forest model used by us gave an accuracy of about 80% in the testing dataset. Further work to increase accuracy can be done.

Source 1 Source 2

Machine learning course by Andrew Ng. You can prefer other sources for learning basic techniques.

Gourav Sharma, Sagardeep Thakre

We have a given data set of Customer eligibility requirements for loan like Gender, Marital Status, Education, Number of Dependents, Income, Loan Amount, Credit History and others. We will apply machine learning algorithms on data set To predict whether the loan will be approved or not.

Kaggle ,analytic vidhya

NA

Puran Mal Jaluthariya

An application of electronic health record

NA

MIMIC III Dataset

Abhishek

Use different forecasting techniques to forecast load.

Provided link for lstm, you can take it one step further and create a hybrid of lstm and cnn.

Source 1

NA

Preetpal Jindal

You are provided with a handwritten captcha and you have to guess the characters on that. Used convolutional nueral network, open cv and contours.

Many companies like Adobe with MDSR profile generally ask for machine learning project and also give some questions on the ml topic .

You can watch you tube tutorial for CNN .if you are very new to this course you can start with Coursera ml Andrew ng course.

Anshita Vijayvargiya

A recommendation system is built using embeddings for movies and users that predicted ratings by a certain user for particular movies.

You'll get to learn the application of machine learning by building and training your own model.

Coursera | Blogs (towards data science) | optimization courses

Priyanshu Kumar

Implementation of the traditional movie recommendation engines

Get to learn about great deal about how OTT platforms recommendation and youtube recommendation works. You will need to watch numerous videos and read a number of articles online to understand it to the core. Preferably refer to the 2006 netlfix competition.

Source 1

NA

anshuman kumar

This model automates the process of purchasing old cars at reasonable price! Avoids bargaining! Movie Recommender system predicts the ratings that a user would give to a movie he/she has not watched/rated.

The model accuracy is high(~90%) Movie Recommender system: Highly practical, its hybrid system is being used by Netflix, YouTube etc today.

Source 1 Source 2

NA

Chaudhary Satwant Singh

- Analysed and scaled over 30 macroeconomic, political, demographic, and sports-related variables to predict a nation's success in the Olympics. - Used K-Means clustering to gain insights on predictor variables and conducted an in-depth Case Study to draw focus on the most significant success factors for a country. - Applied XGBoost Regressor to train the model and predict the Medal Tally of Tokyo Olympics 2020

NA

NA

Trishit Singla

Work on pos tagging using cross domain and cross task transfer learning

NA

Google it

Arpit mehta

https://docs.google.com/presentation/d/163E1KgP2KlJLB1TinNLNthlFOKrjEoHECpmbuE4sZQY/edit?usp=sharing

NA

Presentation Link Given

Ritwik Ghorui

Neutal Networks was used to detect amd locate fault

NA

Already made microgrid form matlab website

Manas Asati

To classify the given dataset using Random forest classification

NA

Source 1

NA

Utkarsh Singh

Apply two algorithms called Lin UCB, Lin Thomson Sampling mentioned in paper. We have to calculate attraction probability of each movie according to these to algorithms.

NA

Source 1

NA

Shivam Tomar

Rates the cv of candidate by identifying keywords

NA

Source 1

NA

Kshitij Sharma

Tried to get the selective voice amplified and clear from a noisy crowd using a amplifier system and a ml model trained on specific voice Initiative to separate all the sound sources that are present in class and then amplify profs voice. It Is a very tough problem, so we started with signal separation algorithm(ICA) just applied that. It also needs various specifications. But was good learning. It really stands out in your resume. Believe me or not.

We got the separated voice of our choice but not will full accuracy as the part of amplification was not much focused upon.

Just Google the cocktail party problem Google independemt componet analysis-- unsupervised learning problem. Search about google voice, Alexxa etc.

Adarsh Srivastav

To build a bot that walks on a black path on a white background. Turning, speed, stopping automated by machine learning.

Finally have a fully autonomous working vehicle.

Yolo project is similar to it, you only look once project.

Tweets from the pre-election period (March-May) were collected from the twitter archive. The collected tweets were classified into two categories, Hate or Not, for subtask A. Hate tweets were further classified into offensive or hateful for subtask B. Compared various Machine learning algorithms like k-NN, SVM, Logistic regression and Random forest based on F1 score and accuracy on our dataset. Used ensemble method on the ML model to boost the performance of the overall model.

The main aim of the project is to get an overall idea about how several people/parties use social media to manipulate the general public or to spread their propaganda.

Source 1

NA

Bhavik Jain

The dataset contained hourly consumption of the building of span of 2 years.this data is sequential .you need to forcast load variations and find trends and seasonality in data if any.lstm gives best results in most cases.

NA

Code can be found on github.Use medium ,analytic vidhya to study concepts involved.

Divya Sinha

Forecast load with various model ARIMA, SARIMA, LSTM, unstacked LSTM, CNN, CNN LSTM .

Gain the knowledge of Time -series forecasting.

Source 1

NA

Manohar Kumar

Proposed a short term load forecasting model based on deep residual networks focuses on the forecasting of loads from several minutes upto one week into the future. Exposure: Python, Keras ,Residual networks

NA

Source 1

NA

Sanika jain

To predict the energy requirement of localities based on smart meter data of customers. So that production and supply of electricity can be managed accordingly and loss could be minimized.

1 or 2 day load requirement of housholds

kaggle data and notebooks

Chetan Garg

Classifies spam and non spam messages

NA

Coursera deep learning course

Pratyush Amrit

Data of many places at different points of time was organised into different time and space bins so that it can be used to analyse trends and behaviour of the region wrt a particular event

You will learn how we can study the region statistics to understand various trends so that appropriate plans can be made accordingly with regards to infrastructure, management etc. (this project can be used for work in the areas of healthcare, pollution, demographics, public amenities, business strategies and economic development etc.)

ArcGIS tutorials on ESRI website and youtube, several other spatio temporal analysis project

Avani Singh

If u have an interest in doing ML project but sometimes B.Tech project is more specific towards your branch than this would be perfect for you.

In intern if you sitting for ML field you will have one project extra than other that is Btech project. And with this you will learn about ML algo.

Contact me if you want to do this project, I will provide my Jupyter Notebook and pdf submitted.

Arushikha Tayal

Stock price prediction

NA

Kaggle

Ekansh Sharma

1. Text Summarization using NLP libraries in Python. 2. Made Flask Based Web Application for interface. 3. Used OCR to generate summaries from Image. 3. Hosted the web app on Heroku.

NA

Source 1 Source 2

https://github.com/shubhdeep22/Text_summarizer_app

Shubham Kaushik

Classify whether the twit is positive or negative using different classifier

NA

Udemy

Bhupendra singh

An unbeatable tic tac toe where player cannot win against the computer

Game should run without any error, in any case the game should end in either a draw or with machine winning the game

Google

Swati

Using GANs for open-set person re-identification

NA

Source 1

NA

Dhvani Shah

The task is to implement an automatic number plate recognizer in an unconstrained condition that considers occlusion, poor quality of images, and other spatial variations in image data.

Text obtained from images of Number plates

Source 1

NA

Pushker Naresh

a novel formulation titled Federated Deep Q Networks (F-DQN) to perform distributed learning for Deep RL algorithms.

Faster Learning than vanilla methods

Source 1

NA

Nishant Kumar

Ive attached my github account link , I've done tons of projects so would be hard to put all in form , please see them

.

Source 1 Source 2 Source 3 Source 4

.

Ayush Agarwal

Apply neural networks to images for classification/identification

NA

kaggle, coursera(Andrew NG), analytics vidhya

Akarsh Vatsa