dek-receipt-donut-baseline

Baseline Donut model for receipt understanding in CORD format.

Status

Baseline checkpoint (no fine-tuning yet).

This model is provided as an initial reference to validate the end-to-end pipeline (data → training → inference → API).

Base Model

naver-clova-ix/donut-base

Dataset

  • 100 annotated receipt images
  • Annotation format: CORD
  • Dataset currently used for development and pipeline validation

Task

Key Information Extraction (KIE) from retail receipts without OCR.

Metrics

Not evaluated at this stage due to limited dataset size. Target metrics will be reported after fine-tuning on an extended dataset.

Intended Use

  • Pipeline validation
  • Baseline comparison
  • Further fine-tuning

Limitations

This checkpoint is not fine-tuned and does not meet target quality requirements yet.

Downloads last month
15
Safetensors
Model size
0.2B params
Tensor type
I64
·
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train SvetaLana25/dek-receipt-donut-baseline