Projects
|
AutoMeta
An open source tool to extract metadata from the cover page of scanned Electronic Theses and Dissertations automatically.
Applied machine learning model such as Conditional Random Field (CRF), utilized sequence tagging method, and incorporated textual and visual features.
The model achieved 83%--96% F1 score on seven metadata fields: Title, Author, University, Degree, Year, Program, and Advisor.
|
ErrorDetect
Error Detect is an open source tool to detect real word and non-word post OCR errors from scanned Electronic Theses and Dissertations.
Applied machine learning model such as SVM with 5 fold cross validation and followed few huristics.
The model achieved a baseline F1 score 48%--72% using different heuristic approaches: Lexicon Lookup, RegEx, 5 gram Kneser-Ney smoothing with Levenshtein distance.
|
Capstone Project: Basketball Training Machine (Final Report)
Undergraduate Project: Elizabethtown College
Designed and constructed a programmable basketball training machine that can track the user around the half-court.
Oversaw the tracking system especially interfacing Arduino with Bluetooth and assisting in the research of GPS tracking.
Worked on an android application for Bluetooth to interface with Arduino using web tool MIT APP Inventor AI2.
|
Junior Capstone Project: Automated Agriculture Simulations (Final Report)
Undergraduate Project: Elizabethtown College
Worked with the real-world client in France to improve the sensor system; changed the algorithm of the data in the Excel model.
Implemented the idea a random water system for the sensor and improved the algorithm for the sample data.
Wrote code in Python for interfacing the valve of the water pump with Raspberry Pi and PLC.