Sheet Music Watermark Removal & Upscale

STAR Summary

• Trained a UNet model on paired page images for watermark removal.
• Upscaled low-resolution scans (e.g., ~612x792) to print-friendly resolution (e.g., ~1700x2200) with Very Deep Super-Resolution (VDSR).
• Wrapped it in a simple GUI so I could run the full pipeline.
• Typical runs produce cleaner, more readable output in seconds per page (hardware- and batch-dependent).

Project Links

• GitHub
• PDFs
• Notebook

Notes

Designed for low-resolution, watermarked page images; output quality depends on scan quality and staff-line preservation.

Other Projects

• Smart Sentence Retriever NLP Embeddings & Serverless Retrieval
• Chatbot (LoRA + RAG) RAG Chatbot Fine-Tuned with LoRA
• Shape Classifier Demo Handwritten Shape Recognition