EmotSense (Emotion Detection sensor)
• Developed and implemented a cutting-edge Emotion Detection sensor using React, Chalice, JavaScript, HTML, CSS, and Python to provide real-time emotion detection for blind individuals facing others, increasing awareness of their surroundings.
• Collaborated with the team to design eyewear concept to enhance usability and accessibility for blind users when interacting with others face-to-face.
• Utilized AWS Rekognition and Amazon Polly services to analyze emotions in images captured by the sensor with an accuracy rate of over 90%, ensuring a reliable and efficient user experience.
Sentiment Analysis of Amazon Review Dataset
• Utilized VADER and TextBlob lexicon models to conduct sentiment analysis on a dataset of 10,000 Amazon reviews, achieving an accuracy rate of 85% when compared with machine learning algorithms.
• Implemented logistic regression and Naive Bayes algorithms to optimize sentiment analysis process, resulting in a 10% increase in overall accuracy compared to lexicon models alone.
• Collaborated with data science team of 5 students to develop a custom algorithm for sentiment analysis, leading to a 15% improvement in accuracy over pre-existing machine learning models.
SafeCompass (AI based traffic collision predictor)
• Utilized machine learning algorithms in Python to analyze the KSI data set from the Toronto Police Service, resulting in accurate predictions for accident severity when implemented in the AI web app.
• Implemented interactive data visualization using JavaScript and HTML to present findings from the KSI data set, increasing user engagement by 40% compared to traditional static reports.
• Collaborated with team of 5 students to deliver a successful presentation showcasing the AI project at a technology conference, receiving positive feedback.
