Add Neural Metrics card (upstream card preserved below)
Browse files
README.md
CHANGED
|
@@ -3,8 +3,34 @@ license: mit
|
|
| 3 |
size_categories:
|
| 4 |
- 1K<n<10K
|
| 5 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
# OmniAI OCR Benchmark
|
| 7 |
|
| 8 |
A comprehensive benchmark that compares OCR and data extraction capabilities of different multimodal LLMs such as gpt-4o and gemini-2.0, evaluating both text and JSON extraction accuracy.
|
| 9 |
|
| 10 |
-
[**Benchmark Results (Feb 2025)**](https://getomni.ai/ocr-benchmark) | [**Source Code**](https://github.com/getomni-ai/benchmark)
|
|
|
|
|
|
|
|
|
| 3 |
size_categories:
|
| 4 |
- 1K<n<10K
|
| 5 |
---
|
| 6 |
+
<div align="center">
|
| 7 |
+
|
| 8 |
+
# Neural Metrics · How good is your OCR, really?
|
| 9 |
+
|
| 10 |
+
<img src="https://img.shields.io/badge/Neural%20Metrics-document%20extraction-4F46E5?style=for-the-badge" alt="Neural Metrics" />
|
| 11 |
+
<img src="https://img.shields.io/badge/fork%20of-getomni--ai%2Focr--benchmark-2563EB?style=flat-square" alt="fork" />
|
| 12 |
+
|
| 13 |
+
</div>
|
| 14 |
+
|
| 15 |
+
A benchmark of real documents paired with ground-truth JSON, designed to score end-to-end accuracy rather than raw character error rate. That distinction matters: an OCR pass can be 99% correct at the character level and still get the invoice total wrong.
|
| 16 |
+
|
| 17 |
+
**We use it for:** scoring candidate OCR models on the metric that actually pays the bills - regression testing before a model swap.
|
| 18 |
+
|
| 19 |
+
> ### Attribution
|
| 20 |
+
> This is an **unmodified fork** of [`getomni-ai/ocr-benchmark`](https://huggingface.co/datasets/getomni-ai/ocr-benchmark), created by the [Qwen team](https://huggingface.co/Qwen).
|
| 21 |
+
> All weights, files and behaviour are identical to upstream — we rehost it so our experiments stay
|
| 22 |
+
> reproducible and version-pinned. The original license and all credit remain with the Qwen team.
|
| 23 |
+
> If you want the canonical dataset, please use [the original](https://huggingface.co/datasets/getomni-ai/ocr-benchmark).
|
| 24 |
+
|
| 25 |
+
---
|
| 26 |
+
|
| 27 |
+
<details>
|
| 28 |
+
<summary><b>Original dataset card from getomni-ai/ocr-benchmark</b> (click to expand)</summary>
|
| 29 |
+
|
| 30 |
# OmniAI OCR Benchmark
|
| 31 |
|
| 32 |
A comprehensive benchmark that compares OCR and data extraction capabilities of different multimodal LLMs such as gpt-4o and gemini-2.0, evaluating both text and JSON extraction accuracy.
|
| 33 |
|
| 34 |
+
[**Benchmark Results (Feb 2025)**](https://getomni.ai/ocr-benchmark) | [**Source Code**](https://github.com/getomni-ai/benchmark)
|
| 35 |
+
|
| 36 |
+
</details>
|