REPRODUCIBLE FIXTURES

Download the sample corpus.

Thirty fictional, image-only PDF scans cover clear, low-resolution, skewed, cross-page, complex-column, and decimal/negative cases. Compare an extraction with the standard answer.

CC0-1.0 · synthetic values · never submit the pristine originals as scanned benchmark inputs.

Clear scans part 1 (2.2 MB) · Clear scans part 2 (2.2 MB) · Clear scans part 3 (2.2 MB) · Clear scans part 4 (2.2 MB) · Clear scans part 5 (2.2 MB) · Low resolution part 1 (0.2 MB) · Low resolution part 2 (0.2 MB) · Low resolution part 3 (0.2 MB) · Low resolution part 4 (0.3 MB) · Low resolution part 5 (0.2 MB) · Skewed scans part 1 (0.6 MB) · Skewed scans part 2 (0.6 MB) · Skewed scans part 3 (0.6 MB) · Skewed scans part 4 (0.6 MB) · Skewed scans part 5 (0.6 MB) · Cross-page headers part 1 (4.0 MB) · Cross-page headers part 2 (4.0 MB) · Cross-page headers part 3 (4.0 MB) · Cross-page headers part 4 (4.0 MB) · Cross-page headers part 5 (4.0 MB) · Complex columns part 1 (2.2 MB) · Complex columns part 2 (2.2 MB) · Complex columns part 3 (2.2 MB) · Complex columns part 4 (2.2 MB) · Complex columns part 5 (2.2 MB) · Decimals and negatives part 1 (2.2 MB) · Decimals and negatives part 2 (2.2 MB) · Decimals and negatives part 3 (2.2 MB) · Decimals and negatives part 4 (2.2 MB) · Decimals and negatives part 5 (2.2 MB)

Download full manifest JSON

New: PDF.co six-file scan pilot — raw outputs and observed errors