From prioritizing car crashes and spotting Anti-vaxxers to perfecting Pilates poses and measuring our moods, here are ITC’s favorite student final projects for 2021.
The old adage of “Practice Makes Perfect” gets a different interpretation here at Israel Tech Challenge (itc.tech). Throughout the intensive tech training, students face a variety of real-world scenarios and challenges, which help them acclimatize quickly into their post-graduation positions in the field of data science. The final project is a cumulation of 5 months of daily intensive studies (or 10 months of part-time studies) and shows us exactly how Data Science and Machine learning can change our world for the better.
Here are the top five full-cycle projects which our teams of data science fellows students submitted before heading off to the 5-weeks industry project.
Classifying emotions using audio recordings and Python
In this project, the students trained a machine-learning algorithm to detect emotions in voice. Their work included preprocessing audio files, extracting the relevant features, and training a neural network to detect clusters corresponding to emotions such as surprise, happiness, anger, and disgust.
Predicting California traffic collision outcomes
The students created a model that predicts the severity of injuries in traffic collisions, in order to help first responders prioritize their responses. They also examined the importance of different features in order to help understand the most important factors in these collisions, including seat belt usage and time of day.
COVID vaccine tweets
Social media has been a key factor in spreading both accurate information and misinformation regarding the COVID-19 pandemic. In this project, the students created a machine learning model to detect clusters of anti-vaccine tweets in order to help governments proactively intervene and encourage vaccine adoption.
Pilates pose detection
The students created a model for detecting different Pilates poses from a video stream, for use in providing feedback during exercises. The team used a creative method for building a dataset based on video footage from YouTube, augmenting the dataset, and training a deep learning model on this data.
Detecting product similarities by their e-commerce titles and images
E-commerce is a massive industry that simplifies shopping for many products, but users may have trouble sorting through all the relevant information. Automatic product matching helps users compare the same or similar products across different sellers to find better deals, or even to find a product that was photographed in real life. The students created an ensemble method involving deep learning with a Siamese network and both textual and image input, and created a web interface for their final application.
If you’d like to explore whether a career in Data Science in Israel is right for you, please contact us today. We offer both part-time and full-time fellows.