Anirban Chatterjee

Data Science Intern

Algo Labs

May 2025 - Jul 2025

Developed an explainable CV system using Streamlit to automatically verify torn document halves for forensic and legal analysis. Constructed a custom 39-pair dataset and designed a four-stage pipeline integrating contour detection, edge extraction, GICP alignment, and Chamfer/overlap verification. Achieved 92.3% verification accuracy, outperforming traditional methods like SIFT and ICP, with step-by-step visual interpretability for decision transparency. Addressed noise and misalignment issues through adaptive preprocessing; proposed automated orientation correction (Homography/CNN) and multi-fragment document matching for scalability.