Projects03 / 06

ClearScan

Built for the Olof Palme murder investigation archives, ClearScan turns faded, handwritten, and noisy case PDFs into usable text through an offline, CPU-friendly AI/OCR pipeline.

Examples of Olof Palme case archive documents handled by ClearScan

01Designed for Softwerk × LNU AI Society's challenge to digitize a subset of the Olof Palme investigation archives, where faded typewriter text, handwriting, and scan noise defeat standard OCR tools.

02Uses contrast-aware image enhancement and local Qwen3-VL inference to transcribe multi-page PDFs, including Swedish characters, while marking redacted or unreadable passages.

03Runs entirely offline on consumer CPU hardware and writes ordered text files. Our team, OCR Titan, won first place in the low-resource OCR challenge.

PythonQwen3-VLllama.cppOpenCVPyMuPDFDocker