New text detection model

New text detection model detects more text across a range of PDFs. Should improve OCR performance.

Fix bug with model downloads.

What's Changed

• Update to new text detection model by @tarun-menta in #328
• Dev by @VikParuchuri in #329

Full Changelog: v0.12.1...v0.13.0

Improved inline math model

What's Changed

• Update to new inline math model by @tarun-menta in #323
• Update inline math model by @VikParuchuri in #326

Full Changelog: v0.12.0...v0.12.1

Model downloads with S3

Download models with S3

• Improve speed and reliability by downloading models with S3

Misc fixes

• Use opencv headless to avoid GUI dependencies

What's Changed

• Add s3 loading functionality to from_pretrained by @tarun-menta in #320
• Swap to headless opencv by @VikParuchuri in #321

Full Changelog: v0.11.1...v0.12.0

Fix inline detection bug

• Fix streamlit bug
• Fix inline detection bug

v0.11.0

Inline math detection

• Add new inline math detection model and benchmark

Textract OCR benchmark

Benchmark surya against textract as well as google cloud vision. For just english, results look like:

Model	Time per page (s)	Avg Score	English
surya	0.522628	0.983298	0.983298
textract	1.44293	0.947458	0.947458

XLA support

Add support for TPUs. Still fairly slow, but lots of optimizations to be made.

Minor speedups

Refactor inference to get a 5-10% speed boost across all models.

What's Changed

• Add XLA support by @iammosespaulr in #298
• Add Inline Math Detection by @tarun-menta in #297
• Update to new line detection model by @tarun-menta in #305
• Fix merging of inline boxes by drawing textlines in heatmap by @tarun-menta in #309
• XLA improvements by @VikParuchuri in #306
• Update inline math checkpoint by @VikParuchuri in #310
• Misc Line Detection Fixes by @tarun-menta in #313
• Add Textract OCR Benchmark by @tarun-menta in #307
• Inline math model, new text detection model by @VikParuchuri in #312

Full Changelog: v0.10.3...v0.11.0

Fix height issue

Fix an issue where text detection wouldn't resize images properly, leading to bounding boxes in the wrong place in tall images.

Fix pytorch 2.6 bug

Fix bug that caused issues on MPS (Mac) devices when using pytorch 2.6.

Pin pytorch

Pytorch 2.6.0 doesn't work well with some of the models on MPS (Mac), so pinning to the old version.

Add LaTeX OCR model

New OCR model and streamlit app

• Release a new LaTeX OCR model
• Add streamlit app to interactively select and OCR equations

What's Changed

• Improve typing for PolygonBox.bbox by @kevinhu in #291
• Add LaTeX OCR by @VikParuchuri in #292
• Texify by @VikParuchuri in #295
• Final texify version by @VikParuchuri in #296
• Integrate new latex OCR model by @VikParuchuri in #293

New Contributors

• @kevinhu made their first contribution in #291

Full Changelog: v0.9.3...v0.10.0

Fix cli script issue

Fix issue with cli scripts and folders.