GSoC 2024: Eye Tracking Algorithm Optimization Based on Low-Resolution Cameras (Final Project Submission)

[sitamgithub-MSIT]

This pull request represents the culmination of my work during Google Summer of Code 2024, focusing on developing RUXAILAB's web eye tracker project.

Key Achievements:

• Completed End-to-End Integration: Successfully integrated the front end with the back end, enabling a fully functional web eye-tracking system.
• Implemented Model Selection: Provided users with the ability to select from a variety of models, including linear regression, ridge regression, lasso regression, elastic net, Bayesian ridge, and SGD regressor.
• Enhanced Data Visualization: Developed a Streamlit app for comprehensive visualization of raw data and metrics obtained during calibration.
• Extensive Model Testing: Conducted thorough testing of various models, including grid search implementations, to identify optimal model choices and assess performance.
• Refined User Interface: Improved the user interface to include model options, calibration configurations, and informative messages.
• Comprehensive Documentation: Generated detailed documentation for all notebooks and files to ensure clarity and ease of use.

Key Challenges and Solutions:

• Overfitting: Identified and addressed overfitting issues in the Y coordinate regression model.
• Model Selection and Integration: Successfully implemented a model selection mechanism and integrated it seamlessly into the front end.
• Performance Optimization: Investigated and mitigated performance issues related to specific models, such as the time-consuming nature of grid search.
• TensorFlowJS Integration: Resolved challenges in integrating TensorFlowJS models, ensuring proper functionality.
• Data Leakage: Successfully identified and mitigated potential data leakage issues, leading to improved model performance.

Future Work:

• Further Model Exploration: Explore additional models and optimization techniques to enhance the accuracy and performance of the eye tracker.
• Advanced Data Analysis: Develop more sophisticated data analysis tools and visualizations to gain deeper insights from the collected eye-tracking data.
• User Experience Enhancement: Continue improving the user interface to provide a more intuitive and user-friendly experience.

This pull request represents a significant milestone in developing RUXAILAB's web eye tracker. I am confident that the advancements made during this project will pave the way for a more robust, versatile, and user-friendly eye-tracking system.

Please review the code and provide feedback. I am open to suggestions and further collaboration.

cc: @hvini @marcgc21 @KarinePistili