Improved Vision For Deepseek
zyh20041227/improved_vision_for_deepseek · model-adapters
Full-coverage image tiling for DeepSeek Harness vision models, dense-text OCR, and document AI.
★ 4GitHub stars
0Forks
2026-09-01Last updated
TypeScriptLanguage
MITLicense
Key features
- 100% geometric coveragecoverage tiles are audited against every source pixel.
- Overlapping seamstext and shapes that cross a tile edge remain visible in a neighbour.
- Content-aware cutsdocument mode moves horizontal seams toward low-ink areas.
- Dense-region reviewan optional 512×512 detail crop supplements, never replaces, base coverage.
- Bounded batcheslarge tile sets are returned in model-safe batches.
Install command
dsh plugin --profile web add github:zyh20041227/improved_vision_for_deepseek