AKSHAY P

Real-time Face Emotion Detection Android App - A native Android application built with Kotlin that detects and classifies human emotions in real-time using TensorFlow Lite and ML Kit/MediaPipe face detection APIs. Features CameraX integration for live camera feed processing, on-device inference with pre-trained CNN models for emotion classification (happy, sad, angry, neutral, etc.), and face landmark detection. Implements efficient real-time video frame processing with Android SDK and displays emotion predictions with bounding boxes overlaid on detected faces.

Customer Churn Prediction using ML & Streamlit - A machine learning web application built with Python, Streamlit, and scikit-learn that predicts telecom customer churn using a Random Forest Classifier. Features an interactive UI with real-time predictions, powered by Pandas for data processing and Joblib for model serialization. Deployed on Streamlit Community Cloud with complete data preprocessing pipeline using LabelEncoders.

WebQuiz - A real-time quiz administration system built with Python Flask and vanilla JavaScript for college tech events. Features an admin dashboard for quiz control (start/reset) and multi-client participant interface with live status updates via HTTP polling. Implements REST API endpoints for quiz management, uses JSON for data storage, and exports results to Excel (XLSX) using OpenPyXL. Built with responsive HTML/CSS using Material Icons and Google Fonts for modern UI design. Supports simultaneous multiple clients without WebSocket setup. Collaborative project developed by three contributors (Abhishek-max825, Sayeem3051, AKSHAY355-a) for academic tech event demonstrations.

Adult Census Income Prediction - A Python machine learning project using the UCI Adult dataset to predict whether individuals earn over $50K annually. Built with scikit-learn, Pandas, and Jupyter Notebook, featuring multiple classifiers (Random Forest, Logistic Regression, SVM, Decision Trees) achieving 85-87% accuracy. Includes comprehensive EDA with Matplotlib/Seaborn visualizations, handles 14 demographic features (age, education, occupation, marital status), implements data preprocessing with LabelEncoder/OneHotEncoder, addresses imbalanced data, and evaluates models using ROC-AUC curves and confusion matrices.

Deep Facial Emotion Recognition (DeepFER) - A deep learning project built with Python, TensorFlow/Keras, and OpenCV in Jupyter Notebook that classifies seven facial emotions (angry, disgust, fear, happy, sad, surprise, neutral) from the FER2013 dataset. Features a custom CNN architecture trained on 35K grayscale images achieving 70-76% accuracy, saved as best_emotion_model.h5 for deployment. Includes real-time emotion detection via realtime_emotion_detection.py using Haar Cascade face detection with OpenCV webcam integration, data preprocessing with NumPy, model evaluation metrics, and visualization using Matplotlib. Implements data augmentation techniques for improved model performance.

Automobile Price Prediction - A Python regression project using the UCI Automobile dataset to predict car prices based on vehicle specifications (company, body-style, engine-type, horsepower, mileage). Built with scikit-learn, Pandas, and Jupyter Notebook, implementing multiple algorithms (Linear Regression, Random Forest, Decision Trees, Ridge/Lasso, Gradient Boosting) achieving 85-95% R² accuracy. Features EDA with Matplotlib/Seaborn visualizations, data preprocessing with LabelEncoder for categorical variables, StandardScaler normalization, and model evaluation using RMSE, MAE, and R² metrics with NumPy for computations.​