Click on the image to be redirected to project’s repo/hosting. For Projects with embedded videos click on the green text to be redirected to respective repo.
LiftSmart: A Machine Learning Based Approach For Classifying Barbell Exercises using Wearable Sensor Data
I developed a machine learning model that classifies barbell exercises and counts repetitions using accelerometer and gyroscope data. The project involved data preprocessing, feature extraction, and visualization using Python. I trained a model that accurately identifies exercises and counts repetitions, achieving high accuracy. This project demonstrates the potential of wearable sensor data in fitness tracking and exercise analysis, enabling personalized feedback and optimized workout routines.
Styling the Future: A Personalized Fashion Recommendation System for Myntra
I developed a fashion recommendation system using the Myntra dataset, leveraging transfer learning. A pre-trained ResNet model was fine-tuned on the Myntra dataset to extract features from product images.
The ANNOY algorithm efficiently searches for similar images, offering benefits over cosine similarity in scalability and computational complexity. Used by Spotify, ANNOY enables rapid retrieval of similar products, powering personalized fashion recommendations for Myntra customers.
The Movie Matchmaker: MobiFlix - A Recommendation Engine that Brings Users and Movies Together
I developed MobilFix, a recommendation system using cosine similarity and the Bag of Words technique. It integrates the TMDB API for movie searches, recommendations, and detailed information on movie statuses, directors, and actors. Additionally, it features a Rafael Dialogflow-based chatbot to assist developers.
The application was developed using Streamlit, which enabled me to create a comprehensive web app with a user-friendly interface. The initial version also featured MongoDB authentication for secure user management.
Furthermore, I implemented a sentiment analysis model using Naive Bayes, which is currently under development. This feature aims to analyze user reviews and provide insights into movie sentiment.
Duplicate Question Detector: A NLP-based Model for Identifying Similar Questions
I developed an NLP-based model using XGBoost to identify duplicate questions in the Quora dataset. By leveraging the power of XGBoost’s ensemble learning, I achieved an accuracy of 80% in detecting duplicate questions. This project showcases my ability to tackle complex NLP tasks and optimize model performance. The model’s high accuracy demonstrates its potential to improve the efficiency of question-answering systems and reduce redundancy in online forums.
Laptop Price Oracle: An ML Model for Predicting Laptop Prices
A Laptop Price Predictor is needed to predict laptop prices with high accuracy. In the dynamic market of laptops, predicting prices accurately is a challenging task. My solution addresses this by developing a robust regression model that can predict laptop prices with high accuracy to aid both sellers and buyers.
The model, trained on a comprehensive dataset, achieved an impressive accuracy of 89%. I used Google Colab for development and MongoDB for data management. My solution provides insights into the factors influencing laptop prices, serving as a valuable tool for both sellers and buyers in the laptop market.