Document Layout & Script Recreation
End-to-end optical character detection (OCD), script transcription (OCR), and typography-aligned page reconstruction.
1
Select Source Document
Choose an archival document from the annotated Kaithi examples or upload your own scan.
Random Kaithi Sample
Load a randomly selected Kaithi page with annotated boxes and transcriptions.
Custom Image Upload
Upload a PNG or JPG. Image enhancement runs automatically while the original is retained for accurate OCR.
Selected Document
—
Ground Truth Available
Custom Upload
Enhanced via Real-ESRGAN
1
Layout Detection (OCD)
Pending
2
Script Recognition (OCR)
Pending
3
Page Reconstruction
Pending
Section 1
Document Layout & Recognition
Ground Truth Words
—
Detected Words
—
OCD Latency
—
OCR Latency
—
Bounding Box Layout Inspection
Predicted
Layout
Ground Truth
Word-Level Transcription Inspection
25 randomly selected crops comparing model predictions with ground-truth transcriptions.
Section 2
Reconstructed Page Comparison
Open Layout Recreation
Model Prediction
Rendering Open Layout…
Rigid Grid Recreation
Model Prediction
Rendering Rigid Layout…
Open Layout Recreation
Ground Truth
Rendering GT Layout…
Rigid Grid Recreation
Ground Truth
Rendering GT Rigid…
Section 3
Devanagari Transliterated Reconstruction
Transliterated Open Layout
Model Prediction
Rendering Transliterated Layout…
Transliterated Rigid Layout
Model Prediction
Rendering Transliterated Layout…
Transliterated Open Layout
Ground Truth
Rendering GT Transliterated…
Transliterated Rigid Layout
Ground Truth
Rendering GT Transliterated…